Skip to main content

Full text of "Learning Theory Personality Theory And Clinical Research"

See other formats


m< OU_1 60032 >m 


Call No. /$*t. tj. Accession No. 

A/Q / 

Author ' * 

Title < 


Ihis book should be returned on or before the date last marked belowv 

Learning Theory, 

Personality Theory, 


Clinical Research 

Eleven lectures 

given under the auspices of 

the Department of Psychology in 

the College of Arts and Sciences 

of the University of Kentucky 

on March 13 and 14, 1953 

Learning Theory, 

Personality Theory, 

Clinical Research 







All Rights Reserved 

This book or any part thereof must not 
be reproduced in any form without 
the written permission of the publisher. 



In recent years learning theory, personality theory, and clinical re- 
search have been among the most vigorously cultivated fields in Ameri- 
can psychology. Many workers in these fields have tried to integrate 
the results of their labor. Some learning theorists have devoted them- 
selves to problems which seem important also for personality theory 
and clinical psychology. Since the aim of psychotherapy is to produce 
changes, and the study of such changes is the study of learning, some 
clinicians interested in psychotherapy have looked to learning theory 
for help in explaining the development of normal and abnormal per- 
sonalities and for guidance in building up a theory of psychotherapy. 

In order to encourage current tendencies leading toward a closer 
integration of these three branches of psychology, the Department of 
Psychology of the University of Kentucky decided in the autumn of 
1952 to hold a symposium on the relationships among these three 
areas. The original proposal was made by Dr. Robert E. Bills, who 
was made Chairman of a Symposium Committee, the other members 
of which were Drs. Lysle W. Croft, P. L. Mellenbruch, Robert D. 
North, and Harold Webster. The University administration gave its 
cordial support to the project, and a number of distinguished psychol- 
ogists accepted invitations to contribute lectures. 

The papers here published were presented at the Symposium on 
March 13- and 14, 1953,- before members of the University community 
and a considerable number of visitors. 

We wish to express our thanks to the President of the University, 
Dr. Herman L. Donovan, for his interest and for his address of wel- 
come at the opening meeting. We also wish to thank Dr. James S. 
Calvin, Dr. Betsy W. Estes, Dr. Charles F. Diehl, Dr. Ernest Meyers, 
Professor Edward Newbury, and Mrs. Lysle W. Croft for their con- 
tributions, made in a variety of ways, to the success of the Symposium. 
Dr. Frank A. Pattie was asked to assume the task of publication of 
the Symposium, and we wish to express our appreciation of his work 
in finding a publisher and in seeing the book through the press. 


Lexington, Kentucky, 
February 12, 1954 





Professor of Psychology, 
Duke University 

Assistant Professor of Psychology, 
University of Louisville 

Assistant Professor of Psychology, 
University of Chicago 

Research Professor of Psychology, 
University of Illinois 

Professor of Psychology, 
University of Wisconsin 

Professor of Psychology, 
University of Michigan 

Research Professor of Psychology, 
University of Illinois 

Professor and Chairman of the 
Department of Psychology, 
State University of New York 
Teachers College, Oswego 

Professor and Head of the 
Department of Psychology, 
State University of Iowa 

Professor of Psychology, 
Ohio State University 

Research Associate and 
Associate Professor, 
Department of Psychology, 
Yale University 



Current Interpretations of Learning Data and Some Recent De- 
velopments in Stimulus-Response Theory, Kenneth W. Spence 1 

Stimulus-Response Theory as Applied to Perception, Delos D. 
Wickens . . . . . 22 

Motivational Forces Underlying Learning, Harry F. Harlow 36 

The Premature Crystallization of Learning Theory, Norman R. F. 
Maier . . . . .54 

Learning and Explanation, Donald K. Adams 66 

Ego Psychology, Cybernetics, and Learning Theory, O. H. 
Mowrer 81 

Personality Structures as Learning and Motivation Patterns A 
Theme for the Integration of Methodologies, Raymond B. 
Cattell . 91 

Prospects and Perspectives in Psychotherapeutic Theory and Re- 
search, John M. Butler . . . .114 

Learning: an Aspect of Personality Development, Donald Snt/gg 129 
"Errors": Theory and Measurement, R. B. Ammons . , 138 

Some Current Research Issues in Clinical Psychology, /. R. 
Wittenborn 148 

Index .161 


Current Interpretations of 
Learning Data and Some 
Recent Developments in 
Stimulus-Response Theory 


In this opening address of the Symposium on Relationships between 
Learning Theory, Personality Theory, and Clinical Research my assign- 
ment is twofold: first, to discuss briefly the data investigated in learn- 
ing experiments and the theories proposed concerning them, and sec- 
ond, to present some recent developments in my own theoretical posi- 
tion. The limited time available necessitates a highly selective treat- 
ment of both topics. Hence I have rather arbitrarily chosen to empha- 
size certain aspects of them on the basis either of their interest to me 
at the moment or of their seeming relevance to this symposium. 

The first point that I should like to bring up has to do with a 
matter that is of considerable importance in any discussion of learn- 
ing theory, particularly one that occurs in the context of such topics 
as personality theory and clinical psychology. I have reference to 
the importance, indeed the necessity, for us to have clearly before us 
what it is that learning theories are about. What are the phenomena 
that the learning psychologist has taken as the object of his studies 
and about which he has attempted to formulate his theories? 

All of us, of course, are more or less familiar in a general way with 
what is meant by learning phenomena, and we have each actually had 
many first-hand experiences with instances of learning in everyday 
life. It should be noted, however, that these familiar instances do not 
constitute the data with which the learning psychologist has been 
concerned. The body of laws and theories that he has developed has 
*jts origins in a very different set of observations. These observations 
have involved a variety of laboratory arrangements in which such 


2 Kentucky Symposium 

things as the lever-pressing behavior and running speed of the white 
rat, the salivary response of dogs, and the closure of the human eye- 
lid have constituted the observed phenomena. The laws that the 
experimental learning psychologist has discovered and the theories 
that he has formulated with respect to them grew out of these labora- 
tory phenomena and as yet have not been related to instances of learn- 
ing in "real" life. 

Unfortunately, the fact that learning psychologists deliberately 
chose not to concern themselves with "real" life situations has not 
always been realized. Still more unfortunate is the fact that, when 
there has been appreciation of it, considerable misunderstanding of 
the purpose of working with such phenomena has existed. Why, the 
question is frequently asked, do not learning psychologists concen- 
trate their efforts on life situations and use human subjects instead of 
white rats, dogs, and monkeys? 

There are a number of answers to such complaints, the complete 
consideration of which, however, would take us too far astray from 
our main concern. My comments will be confined to just two points. 
(1) It should be noted that there are psychologists who do concern 
themselves with learning behavior as it occurs in everyday life. Such 
phenomena have been and are being studied by a number of different 
psychologists, e.g., by the child psychologist, the educational and 
social psychologists, and the clinical psychologist. Presumably what- 
ever knowledge can be gained by such more or less naturalistic, un- 
controlled observation is being obtained. That such studies have 
not led us very far in the attainment of precise laws of even the low- 
est-order type is pretty well known to everyone, and it is this fact 
that brings me to my second point. (2) From the many examples in 
the history of the development of knowledge in other fields of science 
we have come, or at least we should have come, to appreciate that the 
first objective of the scientist, assuming that he has developed ade- 
quate techniques of observation, is to discover the low-order, em- 
pirical laws holding among the particular variables specified by his 
measuring techniques. Both the physical and biological scientists 
have demonstrated over and over the importance of experimental 
control and systematic variation of the combinations of factors in the 
situation under observation if these laws are to be discovered. Fur- 
thermore, these scientists did not hesitate, as some psychologists have, 
to arrange unreal (i.e., unworld-like) situations in order to achieve 
this goal. Although nature may abhor a vacuum, fortunately the ex- 
perimental physicist did not. Likewise, the biologist never troubled 

Spence 3 

himself about the fact that the isolated piece of tissue in his test tube 
was unlike anything in real life. It may be noted, moreover, that the 
laws such artificial arrangements helped to discover and formulate 
subsequently were found to provide for the explanation of events that 
do occur under natural conditions. 

It is in this tradition that the experimental learning psychologist 
has worked. Interested in providing the same kind of knowledge 
about behavior phenomena that scientists in other fields have devel- 
oped, he has proceeded in a comparable manner to arrange for the 
successive isolation of various aspects of learning phenomena so that 
their interrelations might be ascertained. Thus in the case of the 
individual organism he has attempted in some of his experiments to 
eliminate the operation of certain not-too-well-known and hence un- 
controllable processes that we refer to as thinking or reasoning. One 
manner of doing this has been to use non-articulate organisms such 
as the white rat. Another has been to employ situations such as the 
memory drum set-up in which the temporal sequence of stimuli and 
responses is such as to preclude the operation of these unobservable 
processes. In a similar manner the learning psychologist has at- 
tempted to isolate and control the various aspects of the environment 
eliciting and following the responses of the organism. Finally, con- 
trol and manipulation of the motivational factors have also been an 
important part of the learning psychologist's experimental efforts. 

As the result of these controlled laboratory studies a number of 
important behavioral and environmental variables have been identi- 
fied, and at least a good beginning has been made in discovering and 
formulating the type of lawful relations among them that scientists 
seek. For the most part, it is true, these laws are of the low-order 
type involving specific response variables in the several specific ex- 
perimental situations (classical conditioning, discrimination learning, 
paired associate learning, etc.). As yet only a start has been made 
toward the development of the type of higher-order generalizations 
(i.e., theories) that serve to integrate or unify the laws specific to each 
experimental situation. 

At this point we need to take a little closer look at the types of laws, 
hypotheses, and theories that the experimental studies of the learning 
psychologist have produced so far. The laws provided by the differ- 
ent kinds of learning experiments consist in a set of empirical functions 
which relate the various response measures to a number of manipulable 
environmental variables. The following represent some of these in- 

4 Kentucky Symposium 

vestigated relationships for one response measure (e.g., frequency of 
response) in one experimental situation (classical conditioning): 

1. R = /(Number of trials, N). 

2. R = /(Intensity of the conditioned stimulus, S<?). 

3. R = /(Intensity of the unconditioned stimulus, St/). 

4. R = /(Time Jntervaljbetween Sc and $[/ Ts^sJ- 

5. R = /(Timejbetween successive trials, TR). 

6. R = /(Amount of work involved in R 9 W). 
R = f(N,S c , *SV, r Wil , T R , W, etc.). 

The so-called learning function (No. 1 in the list) is one of these 
laws. Although it is the function in which learning psychologists 
have been most interested, the relations of performance measures to 
such other variables as time of food deprivation (T D ), delay of re- 
ward (T0), and intensity of the UCS (S^) are also important for a 
complete account of the behavior of the subject in the situation. 

A number of points concerning these empirical laws are of interest 
to us. First is the experimental fact that a law involving a particular 
response measure may be very different from one experimental situa- 
tion to another. For example, the percentage of correct responses as 
a function of successive blocks of training trials is negatively acceler- 
ated in most simple T-maze studies, whereas the same measure shows 
a period of initial acceleration in difficult discrimination situations. 
Likewise, the learning functions for different response measures in 
the same situation may take very different forms. Thus the frequency 
measure typically shows an S-shaped curve in classical conditioning, 
whereas resistance to extinction is usually a negatively accelerated 
function. In other words, the laws found in our studies are to a very 
considerable extent highly specific to each experimental situation and 
to each response measure. We shall return later to a discussion of 
the implication of this fact for learning theory. 

A second point in connection with these empirical laws of behavior 
in learning situations is that, in addition to being interested in the 
relation holding between each response measure and each inde- 
pendently manipulable environmental variable, the psychologist is 
also very much concerned with the problem of how these latter vari- 
ables combine or interact with each other to determine response 
strength. In other words, he is interested in determining the precise 
nature of the final complex function for each response measure in 
each experimental situation. 

Spence 5 

With these two aspects of the laws of behavior in learning situa- 
tions before us, let us now turn to a brief review of the kinds of 
theories that have come out of these learning experiments. As we 
shall see, there are a number of very different things that are called 
theories, and theories serve a number of different purposes. 

One of the more prominent kinds of learning theory has been the 
intervening-variable type introduced by Tolman [26, 27]. Conceiv- 
ing one of the main tasks of the learning psychologist at the present 
time to be that of discovering and formulating the precise nature of 
such complex laws as that shown at the bottom of the list on p. 4, 
Tolman proposed that it was necessary to introduce theoretical con- 
structs or, as he termed them, intervening variables, in order to do 
so. He believed that the complicated nature of the function relating 
a response variable to its several determining variables made it nec- 
essary to proceed by conceiving of the function as being analyzed 
into successive sets of simpler component functions. According to 
Tolman, these component functions begin by defining a set of inter- 
vening variables in terms of the independent variables. Further in- 
tervening variables are then introduced by stating them as functions 
of the first set, until finally the dependent behavior variable is postu- 
lated as some function of one or more of the intervening variables. 

Actually Tolman never got around to demonstrating how such 
theoretical constructs provided us with these complex laws, and more 
recently he himself seems to have questioned their usefulness [28]. 
While it may or may not be a fact that the discovery of the precise 
nature of such complex functions is facilitated by this intervening- 
variable type of theorizing, attention should be called to the fact 
that it is entirely possible to ascertain them by purely experimental 
means. Thus one could proceed to study the response variable as a 
joint function of various combinations of the independent variables 
by means of the factorial type of experiment. An instance of this 
kind of experiment is the well-known Perm- Williams study [14], 
which showed that the variables, time of deprivation and number of 
trials, combine in a multiplicative function to determine response 
strength. We have recently obtained data in the Iowa laboratory 
which show that the UCS in classical conditioning and number of trials 
likewise combine in a multiplicative manner to determine response 
strength. The main point here, however, is that it is not necessary 
to introduce intervening variables to be able to formulate such com- 
plex laws. The problem is, rather, an experimental one, one that 
we ought to be working at more than we are. 

6 Kentucky Symposium 

A second type of theory that has also employed the intervening- 
variable type of construct is represented by the work of Hull [9, 10, 
11]. The primary concern of this type of theory, as I have pointed 
out in a number of previous discussions [17, 18], has been to build 
a theoretical structure or model that would serve to derive and hence 
interrelate the specific laws found in the different learning situations. 
In his book Principles of behavior Hull began with the empirical 
findings from two of the simplest learning situationsclassical and in- 
strumental conditioning. On the basis of these data he proposed a 
hypothetical system of laws involving intervening variables and ex- 
perimental variables that would provide for the derivation of the 
different laws for each response measure in these two situations. In 
its initial formulation such a theoretical structure is, it is true, purely 
ad hoc. The theorist makes those particular assumptions that will 
lead to the derivation of the known empirical laws. Once formulated 
on the basis of this initial set of data, however, the theoretical model 
may be tested in terms of the degree to which it is able to derive 
lawful relations appearing in new data, whether from the same situa- 
tion or from other learning situations, e.g., in discrimination learning, 
maze learning, and rote serial learning. Hull's posthumously pub- 
lished book, A behavior system, represents an attempt to apply his 
theoretical model to new, more complex learning situations. For a 
number of years the writer and his students have been engaged in a 
similar application of it to the phenomena of discrimination and 
simple selective learning [5, 19, 24]. Other theories that are essen- 
tially similar, in principle, to this type are those of Thurstone [25], 
Gulliksen [6], Pitts [15], and the more recent formulations of Estes 
[3] and Bush and Mosteller [1]. 

A third class of learning theories is represented by the mathemati- 
cal formulations of Rashevsky and his students [8, 16]. These theories, 
like those of the previous group, also attempt to integrate the laws 
from the different learning situations into a more comprehensive 
system of knowledge. They differ, however, in the origin of their 
mathematical assumptions. Instead of beginning with the data from 
one or another type of learning situation they start with certain 
knowledge from neurophysiology. By means of further assumptions 
as to the neurological circuits involved in the different learning situa- 
tions they then derive rational equations representative of empirical 
relationships to be expected in learning data. This type of theory is 
perhaps of even less interest to the psychologist concerned with the 
application of learning theory to phenomena of everyday life than 

Spence 7 

the other theories we have been discussing. Its main contribution is 
the integration it promises to provide of the laws of neurophysiology 
with the behavioral laws found in the learning psychologist's labora- 

In a fourth and final class of learning theories I have lumped a 
variety of widely different kinds of theorizing that do not fall into the 
first three groups. For the most part these are limited, specific hy- 
potheses rather than the more comprehensive type of theory that 
attempts to integrate different areas of laws. These hypotheses vary 
from being mere guesses as to the role some previously unsuspected 
variable plays in a particular experimental set-up, through specula- 
tions as to the physiological mechanism underlying some bit of be- 
havior, to extrapolations of specific laws discovered in controlled 
experimental situations to analogous situations in everyday life. 
Miller and Bollard in their books Social learning and imitation and 
Personality and psychotherapy provide many examples of this sort 
of hypothesizing arid in so doing have made a start, at least, in tying 
together the phenomena of everyday learning and the more precise 
laws discovered in the laboratory. Instances of such limited hy- 
potheses in experimental learning data are the drive-reduction hy- 
pothesis of reinforcement and the hypothetical role of the fractional 
anticipatory goal response in mediating the appropriate response in 
latent learning experiments. 

This concludes the first part of my assignment. In summary, I have 
emphasized the point that the data heretofore investigated by the 
theory-oriented learning psychologist are, for the most part, quite 
different from the observations of learning in everyday life. They 
consist, as we have seen, in observations obtained under controlled 
laboratory conditions that often depart radically from those of every- 
day life. The main characteristics of the laws and the different types 
of theories that these experimental studies have led to were outlined 
briefly, and the need for such an approach was indicated if the ab- 
stract kind of knowledge that the scientist seeks is to be found. 

One final comment before leaving this discussion is concerned with 
the question as to whether the theories developed from these labora- 
tory experiments, when sufficiently elaborated, will ever provide for 
the explanation and control of behavior in real life situations. No 
definitive answer, of course, can be given at the present time, for as 
yet none of them is sufficiently abstract or complete to account even 
for all the laboratory findings. One can only point to the success 
that such types of theory have had in the physical sciences and as- 

8 Kentucky Symposium 

sume that the same developments will occur in psychology. It is 
perhaps worth while for a few psychologists, at least, to have this 
faith, if for no other reason than to provide a control experiment to 
test the beliefs of the many who know on a priori grounds that psy- 
chology is different in this respect from other sciences. 


In turning to the consideration of some recent developments in my 
own theoretical notions concerning learning I should, perhaps, begin 
by identifying my position in terms of the different classes of theories 
mentioned earlier. As was indicated in my discussion of Hull's the- 
orizing, I have for a number of years been interested in extending the 
quantitative aspects of this type of theory to more complex learning 
phenomena, particularly discrimination learning. Through the me- 
dium of doctoral theses we have also been engaged in checking and 
modifying, where necessary, the basic theoretical structure as de- 
veloped within the context of classical and instrumental conditioning. 

This position, as you know, is typically referred to as the S-R re- 
inforcement theory, and as such it is usually contrasted with such S-S 
contiguity theories as that of Tolman and even such S-R contiguity 
positions as that represented by Guthrie [7]. Although the issues of 
S-S versus S-R and reinforcement versus contiguity are genuine 
enough ones, they do not in my opinion warrant the amount of at- 
tention that has been accorded them. Furthermore, from the point 
of view of the development of a quantitative theory that will serve 
to integrate the laws found in different learning situations they are 
of no immediate importance. In other words, such quantitative the- 
orizing is, or can be, entirely independent of these issues. 

Although the concept of S-R tendency will be used in the ensuing 
presentation, no bias is intended as to whether the hypothetical learn- 
ing change is a connection between two sensory elements of the nerv- 
ous system or between afferent and efferent units. The essential mean- 
ing is that of a quantitative value that changes in some specified amount 
as the result of certain experimental operations. Likewise the term 
"reinforcement" will be used but only in the general sense that the 
stimulus effects or consequences of a response in some way or other 
are related to the changing value of the S-R tendency. 

We shall have time to discuss only two recent developments in 
this quantitative S-R theory. The first that we shall consider is not, 
strictly speaking, a new development, but represents rather an emen- 
dation of one portion of Hull's basic theoretical formulation that seems 



to have escaped the attention of even his most ardent "critics. Figure 
1 presents the variables, experimental and hypothetical, that are 
involved in the theory developed for classical and instrumental con- 
ditioning. At the left are represented the experimental variables that 
have been found to affect response strength (S c , intensity of the CS; 
T Sc -8 u > time interval between onsets of CS and UCS; TO, time of delay 
of reward; N, number of reinforced trials, etc.). In the middle are 




R P = normal prob. f(E >L) 
R T =aE~ b + c 
R N =aE-j 

Fig. 1. Diagram showing relation between the experimental variables and the 
intervening variables in Hull's theory of simple learning phenomena. 

the intervening theoretical variables, and at the right are the de- 
pendent response variables. The part of this theoretical structure that 
I wish to discuss now is the last set of relations (represented by the 
broken arrows) at the extreme right of the figure. 

In the Principles of behavior Hull called attention to the importance 
of anchoring the intervening theoretical constructs both to the ante- 
cedent environmental conditions and to the consequent response 
events. The latter he proposed to accomplish by means of postulates 
relating the different response measures to one or another of the rele- 
vant theoretical constructs. Thus, in the case of the three response 
measures we shall be considering, he made the specific assumptions 
shown at the top right of Fig. 1. 

1. The probability of occurrence of a response (R P ) is a normal, 
integral function of the amount that the effective excitatory potential 
(E) exceeds the reaction threshold (L). 

10 Kentucky Symposium 

2. The latency of response (Rr) is a decreasing hyperbolic function 

of the momentary effective excitatory potential (E). 

3. The number of trials required to produce extinction (R N ) is a 
simple linear increasing function of the effective excitatory potential 

Actually these postulated relations are entirely unnecessary, for it 
is possible to derive relations between these response measures and 
one or another of the intervening theoretical constructs on the basis 
of already existing postulates. Thus it may be shown that the func- 

Fig. 2. Diagram showing relations between intervening variables, effective excita- 
tory potential (U), oscillatory inhibition (O), and reaction threshold (L). 

tion relating probability of response (Rp) to E may be derived from 
the assumptions already made concerning oscillatory inhibitory po- 
tential (O) and the response threshold (L). This derivation may be 
shown graphically by means of Fig. 2. In this graph a linear increase 
in E is represented by the line OY; L represents the threshold value 
that a particular momentary value of E must exceed for a response 
to occur. The upended normal distributions represent the frequency 
of occurrence of the oscillatory O values that are subtracted from 
effective excitatory potential (E) to give the momentary effective 

excitatory values (E). The shaded area in each of these distribu- 
tions, then, represents the probability that the momentary effective 
excitatory potential will on a particular trial be greater than the 
threshold value, L. But this probability value, which is designated 

Spence 11 

as p(E > L), also gives the probability of a response occurrence, for 
in the postulate concerning L it is assumed that a response occurs 
only when the momentary effective excitatory potential is greater 
than L. That is, R p has the same value as p(E > L). If now we 
were to plot the areas of the shaded portions of these distributions 
for successive values of (E L) we would obtain a normal integral 
function. There is no need, then, to make a special postulate, as Hull 
did, for R P is necessarily a normal integral function of (E L) by 
virtue of the previous assumptions made concerning O and L. 

Turning now to the latency measure, we can also derive its relation 
to momentary effective excitatory potential instead of making an 
arbitrary postulate as Hull did. The derivation is too lengthy and in- 
volved to give here. Suffice it to say that it involves assigning a value, 
h, to represent the average duration of a momentary O. We need 
then only to be able to calculate the mean expected number of suc- 
cessive momentary O values that will occur before one sufficiently 
small to produce a superthreshold momentary excitatory potential 
results. This value, in turn, may be shown to be a function of the 
probability of getting such an O value. The outcome of the deriva- 
tion is shown in the following equation, in which t is the average 
latency per trial and h is the average duration of a momentary O: 

t = 

p(E > L) 

If we take the reciprocal of t we have the relation between a meas- 
ure of speed of getting into action (S) and the momentary effective 
excitatory potential as follows: 

> L) 

One of the interesting implications of this function is that if we 
now were to plot S against N 9 the number of trials, we should expect 
to get the same shape function that we would get with the probability 
of response measure. Figure 3 shows plots of this hypothetical prob- 

ability value [p(^>L)]asa function of N for a number of different 
values of the parameters that determine the level to which E will 
grow. Those of you who are familiar with curves of learning em- 
ploying such a speed measure will recognize that they conform closely 
to these theoretical curves. 


Kentucky Symposium 

As in the case of response probability and latency it may also be 
shown that there was no need for Hull to make a special postulate 
concerning the relation between the third response measure num- 
ber of trials to extinction (R N ) and the intervening theory. In this 
instance assumptions already made concerning the development and 
dissipation of inhibitory potential (I) lead to definite implications as 
to the relation between R N and effective excitatory potential (E). 
The relationship may be shown to depend upon whether massed or 



35 45 55 65 75 85 95 105 
Trials (N) 

Fig. 3. Family of theoretically derived curves of the proportion of superthreshold 

momentary effective excitatory potentials [p(E> L)] as a function of number of 

training trials for different growth curves of excitatory potential (E). 

distributed conditions of practice obtain. I shall not go further into 
the nature of the implied relations, as my primary purpose has been 
to bring out the point that it was unnecessary for Hull to make this 
final set of assumptions in the case of these three measures. 

Although I have not had an opportunity to examine thoroughly the 
more recent theoretical formulations of Hull, including that presented 
in the first chapter of his new book, A Behavior system, it is my im- 
pression that the same difficulties exist in them. Furthermore, Hull 
has made a number of changes in this portion of his theory, particu- 
larly in his conception of the oscillatory potential, O, which, in my 
opinion, are not for the best. For example, I would question the 
change from his original conception that the dispersion of O is in- 
variable throughout learning to the conception that O begins with 
zero dispersion and increases as a growth function to a maximum dis- 

Spence 13 

persion in the first eight or nine trials. The reason that Hull gives 
for this change is that, once the response in instrumental learning has 
risen above the reaction threshold and has been reinforced, it rela- 
tively infrequently fails to occur during the subsequent learning. Just 
why Hull thought this fact to be contrary to his original conception 
is difficult to understand, for obviously in instrumental learning it is 
just a matter of waiting for the oscillatory O value to become suffi- 
ciently small to produce a momentary superthreshold excitatory 
potential. This happens in instrumental learning because the experi- 
menter usually does wait for the response to occur. It does not 

(V x K x D) x H = E 

(V + K + D) x H = E 



Fig. 4. Diagram showing relation between motivational (V, K, D) and learning 

variables (H). 

happen in classical conditioning because the UCS does not wait for 
the conditioned response to occur. There are other reasons why I 
think the older conception is superior to the new, including some 
experimental evidence on the shape of frequency curves of classical 
conditioning, but time will not permit their discussion here. 

I should like to direct your attention now to a second aspect of our 
theory in which there have been a number of new developments. 
Figure 4 presents that portion of our theoretical structure that deals 
with the motivational factors V, K, and D and their relations to the 
learning factor, H. In this diagram the experimentally manipulable 
variables are shown at the top of the figure, the intervening variables 
inside the rectangle, and the response variable at the bottom right of 
the figure. The arrows indicate which experimental variables are as- 
sumed to determine each of the intervening ones. 

Equation 1 in the diagram represents the most recent Hull con- 
ception as to how the intervening motivational variables V, K, and D 
interact or combine with each other and with the learning factor H 
to determine E and hence response strength. Thus, on the basis of 

14 Kentucky Symposium 

the available evidence in classical and instrumental conditioning, 
Hull assumed that each of the three motivational factors combined 
with H in a multiplicative fashion to determine E. Lacking evidence 
as to how V, K, and D combined with each other, Hull made the ar- 
bitrary working hypothesis that they also combined multiplicatively 
with each other. 

Now I should like to call attention once more to the fact that Hull 
developed this theory in a purely ad hoc fashion to fit the known facts 
of these simple conditioning studies. Whether this motivational the- 
ory is adequate for more complex learning situations remains, of 
course, a question. Obviously, the manner in which such a question 
should be decided is to derive the implications of the theory for each 
complex learning situation and then check them against the experi- 
mental facts. Although this procedure seems obvious, as I have said, 
it has not always been followed. In its stead the flat assumption 
seems to have been made by some of our critics that an increase in 
drive strength always leads to a higher level of performance in more 
complex situations just as it does in simple learning. When this 
assumed result fails to occur, the theory is denounced and the futility 
of attempting to explain more complex behavior in terms of theoretical 
constructs from simpler situations is proclaimed. 

The most charitable interpretation that can be made of this misuse 
of theory is that it is due to a failure to appreciate that the application 
of a theory to any particular situation involves not only a considera- 
tion of the laws or hypothetical relations postulated in the theory, but 
also what are referred to as the initial or boundary conditions of the 
situation. The logical implications of a theory are a joint function of 
the postulated relations or laws and the particular combination of 
conditions or variables operating in the situation. As I will now at- 
tempt to show by means of some examples, the implications of our 
theory as to the effect of varying drive strength in complex learning 
situations are not what has been so naively assumed. The implica- 
tions differ, as we shall see, from one situation to another, and many 
other factors in addition to the manipulation of the strength of a 
single particular need must be taken into account. 

One of the most important of the factors determining the effects 
of drive variation on performance in learning situations is the nature 
and complexity of the initial response hierarchy. If the learning task 
is a simple one in which there are no competing responses but only a 
single S-R tendency as in classical conditioning, or a set of more or 
less isolated single S-R tendencies as may be arranged in paired- 

Spence 15 

associate learning, a high drive level should be expected to lead to 
a relatively high level of responding. Thus in classical conditioning 
there is, by virtue of the control of the stimulus conditions, a single 
highly dominant response. As training proceeds, the CS acquires 
habit strength (H) for this response. The excitatory strength (E) of 
the CS to evoke the response depends, it is assumed, on the product 
of this H and the level of D, i.e., E = f(H X D). The higher the 
value of D, the greater E will be and hence the greater will be the 
response strength. In more complex learning situations involving a 
hierarchy of competing responses, however, the effect of drive-strength 
variation will depend upon a number of factors the influence of which 
must be evaluated in each specific situation. 

Ideally, the instrumental learning situation is also one in which 
there is but a single S-R tendency, and if this were so the same pre- 
diction could unqualifiedly be made for it as for classical conditioning. 
In actual practice, however, instrumental learning situations are limit- 
ing cases of trial and error learning in which there is an initial hier- 
archy of S-R tendencies. Although the experimenter usually chooses 
the strongest response in the hierarchy to reward, or attempts by 
various means to make it the strongest, the presence of the other com- 
peting responses needs to be considered. I shall not attempt to treat 
this situation in detail here, as it is much more complicated than has 
been realized. It is perhaps sufficient to state that after considerable 
training the habit strength of the reinforced response is so much 
greater than that of the competing responses that the latter have 
little or no effect. At this stage it may be predicted that the response 
speed would be directly related to drive level, as in classical condi- 

Let us now consider a selective learning situation in which there 
are two competing responses of about equal strength at the begin- 
ning of training. A very simple situation of this type is one in which 
we have two alternative responses both of which lead to a reward, 
but one of which involves a longer delay of the reward than the other 
(e.g., pressing one of two bars, entering a left or a right alley). We 
recently employed such a situation in an experiment to test an alterna- 
tive hypothesis to that of Hull's as to how the motivational factors K 
and D combine. This hypothesis is shown in the second equation 
of Fig. 4. It will be seen that the hypothesis assumes K and D com- 
bine in some additive manner rather than multiplicatively as assumed 
arbitrarily by Hull. We .were interested in comparing the perform- 
ance of two groups of white rats under different hunger levels in 

16 Kentucky Symposium 

such a situation where the number of responses was kept equal by 
means of forced trials. One of the responses involved a delay in re- 
ward of 1 second, the other a delay of 5 seconds. The deprivation 
times were 3 hours and 22 hours. The prediction for the additive 
assumption is that there will be no difference in the per cent choice 
of the short-delay response over the long-delay under the two levels 
of drive strength. The derivation may be shown algebraically as 

E s = H(D 

E L = H(D + K L ) 
E S -E L = H(D + K s ) - H(D + K L ) 

= H(K a - K L ) 

Thus it will be seen that the difference between the excitatory po- 
tentials of the short- and long-delay responses and hence the per- 
centage choice of one over the other will be a function of H ( i.e., the 
number of reinforced trials on each response) and K (i.e., time of 
delay of reward) but not a function of D (i.e., time of deprivation). 

The implication of Hull's multiplicative assumption for the same 
experimental situation, on the other hand, may be shown to be just 
the opposite, i.e., that performance will be a function of D. The out- 
come of the experiment, although not entirely clear cut, tended to 
favor the additive assumption, for a significant difference between the 
two drive groups failed to appear. My own expectation is that ap- 
propriate experimentation will demonstrate that all three motivational 
factors will be found to combine with each other in an additive man- 
ner as suggested in the second equation of Fig. 4. 

Let us turn now to another competing-response situation one that 
again involves two alternative responses in the hierarchy, such as a 
left and a right alley in a simple T-maze. In this situation one of the 
responses leads to reward; the other to an empty goal box with no 
opportunity to correct. Assuming the responses are equal in strength 
to begin with, or that the correct one has somewhat greater habit 
strength than the wrong one, it may be predicted from our theory 
(either Hull's or my version) that a higher drive level will lead to a 
higher level of performance. The derivation is as follows: 

E+ = J5T+ X D 
E- = BL. X D 

Spence 17 

E+ - E_ = (H+ XD)- (tf _ X D) 
= D(H+ - ff_) 

Here it will be seen that a difference in level of D does determine 
the magnitude of the difference in the excitatory potentials of the 
two competing responses and hence the probability of choice of one 
over the other. Quite in contrast to the previous situation, then, the 
additive hypothesis does imply a difference in performance level un- 
der different drive levels in this non-correction situation. 

Now, as some of you may know, there have been a number of ex- 
perimental studies of the effect of drive level on performance in this 
type of selective learning situation, and the findings have been highly 
conflicting. Obviously there must still be some other factor operative 
in these experiments, differential variation of which has produced 
these conflicting findings. What could such a factor be? One pos- 
sibility that is suggested by our theory has to do with the relative 
habit strengths of the two competing responses at the start of the 
experiment. You will recall that one of the conditions involved in 
our prediction was that, initially, either the responses were equal in 
habit strength or the correct one was slightly stronger. This is an 
extremely important condition, for the implication in which the in- 
correct response is initially stronger in habit strength than the correct 
(reinforced) response is just the oppositenamely, that the high-drive 
group would start out at a relatively lower level of correct choice 
and hence would make more errors than the low-drive group. 

If now the experimenter should throw together into the same group 
subjects with opposite initial response preferences, we would have 
two opposing effects that would tend to cancel each other and thus 
result in no difference in performance under the different drive levels. 
If the experimenter should have only subjects who have no initial 
preferences or only slight pieference for the correct response, a dif- 
ference under varying drive levels in favor of the high-drive group 
would be obtained. Should the experimenter have only subjects with 
strong initial preferences and should he train them against their 
preference, the low-drive group should perform the better, at least 
in the earlier stages of the learning. 

The predictions just made, it should be noted, were in terms of the 
per cent of choice of one response over the other. So far as this fre- 
quency or probability measure is concerned, we can ignore other 
possible superthreshold response tendencies that undoubtedly are 
present. We simply confine our scoring to noting the occurrence of 

18 Kentucky Symposium 

one or the other of these two selected responses. But the situation is 
very different in the case of a time measure, e.g., time required to 
make either response. Particularly would this be so if there were 
other response tendencies in the hierarchy (such as trying to climb 
out of the apparatus) that were stronger than those of approaching 
and entering either of the alleys. In this event it may be shown that 
such a time measure should be longer for high drive levels than for 
low drive levels. That is, the animals would be expected to take 
longer to make either of the responses under the high-drive condi- 
tions than under the low-drive conditions. 

I hope that these considerations have given you a better apprecia- 
tion of some of the problems that are involved in making such 
predictions from theory and that they will convince you of the 
importance of the point I made earlier: namely, that the logical im- 
plications of a theory are always a joint function of the postulated 
relations or laws and the so-called initial conditions. 

Let me finish by just mentioning briefly one or two further con- 
siderations. The matter of the type of measure used is of particular 
importance for the problem-box type of situation in which the meas- 
ure most usually employed is the time taken to make the correct 
response. Such learning situations typically involve a large number 
of responses in the initial response hierarchy, with the correct goal- 
attaining one being relatively low in habit strength. Depending on a 
number of factors, such as the relative position of the correct response 
in the hierarchy, the magnitude of the initial habit-strength differences 
among the alternative responses, and the number of responses of 
fairly weak habit strength, different predictions as to the effects of 
varying drive strengths would be made. Since these initial values are 
usually unknown it is impossible to use them in making precise tests 
of our theory. As a consequence we have been forced to design new 
complex learning situations in which the initial response hierarchies 
can be determined in terms of experiences prior to or during the 

An interesting variant of this latter procedure is the series of moti- 
vational experiments we have been conducting at Iowa over the past 
few years. Employing human subjects, these studies have attempted 
to manipulate drive level by selecting two groups in terms of their 
extreme scores on a test of manifest anxiety. Those of you who are 
familiar with these experiments will recall that we found that sub- 
jects at the high (anxiety) end of the scale exhibited a significantly 
higher level of performance in classical conditioning than did sub- 

Spence 19 

jects from the low (non-anxiety) end of the scale [20, 21, 22]. Inter- 
preting the high end of the scale as reflecting high drive level and 
the low end as reflecting low drive level, it was then predicted that 
the opposite finding, i.e., superior performance on the part of non- 
anxious (low-drive-level) subjects, would be found in more complex 
learning situations in which there was a high incidence of strong, in- 
correct, competing responses. This prediction has been confirmed in 
three separate experiments involving rote serial learning [13], a ver- 
bal maze [23], and a stylus maze [4]. The stylus-maze experiment 
is of particular interest because of the fact that the same subjects 
were run in an eyelid conditioning experiment. Whereas the anxious 
subjects were superior in the conditioning situation, the opposite was 
the case in the maze experiment. 

We have just recently extended this research to paired-associate 
learning. We constructed one list of words in which every effort was 
made to make the correct S-R tendencies relatively strong and to 
minimize the possibility of competing response tendencies. For such 
a list it was predicted that anxious (high-drive) subjects would per- 
form better (make fewer errors) than non-anxious (low-drive) sub- 
jects. In a second list an attempt was made to maximize the number 
and strengths of the competing responses and thus obtain a list on 
which the non-anxious subjects would be superior to the anxious 
subjects. The procedure involved the manipulation of the synonymity 
of the stimulus words and the strength of associative connections of the 
paired words. Preliminary results for only nine subjects in each 
group are shown in Table 1. 

It will be seen that, just as was predicted, the anxious subjects 
performed better on list 1, in which the number and strength of com- 
peting responses were minimized, whereas they were inferior on list 
2, in which competing responses were stronger than the correct re- 

TABLE 1 Mean number of errors made by anxious and non-anxious subjects on 
list 1 (no or low competition) and list 2 (high competition) 

Subjects List 1 List 2 

Anxious (N = 9) 15.2 98.7 

Non-anxious (N = 9) 27.4 79.4 

p <0.01 <0.05 

sponses. It should be realized that different subjects were involved 
in the two lists, but a comparison of the groups on a preliminary 
practice list showed them to be comparable. 

20 Kentucky Symposium 

In concluding I cannot forego pointing out that the type of additive 
hypothesis that I have been employinga curvilinear one along the 
lines of the manner in which Hull assumed habit strengths to summate 
requires that consideration also be given to the strengths of the V 
and K values when making predictions as to what the effects of varia- 
tion in D will be. Very briefly, the predictions we have made hold 
when the values of V and K are relatively low. If these values are 
maximized there will be a general wiping out of the influence of varia- 
tions of D, and performance will tend to be quite independent of 
them. This latter statement holds, of course, even within the factor D 
itself. Thus, as has been shown in a number of instances, different 
strengths of the hunger need do not produce performance differences 
if a strong pain or anxiety condition is also present. 

Significantly enough, it is the global-minded person who, while 
giving full lip service to the principle of taking all the factors into 
account, invariably fails to do so when disproving one of our the- 
oretical predictions. This seeming perverseness is only apparent, 
however; for, lacking any type of theoretical analysis, such a person 
literally doesn't recognize the potential factors and their possible 
action. Although there are dangers in the biases of a theorist, this 
danger is negligible if a genuine effort is exerted to make the hy- 
pothesized relations as specific and precise as possible. Theoretical 
biases thrive only in the vague types of formulation that fail to meet 
this specification. 


1. Bush, R. R., and Mosteller, F., A mathematical model for simple learning, 

Psychol Rev., 1951, 58, 313-323. 

2. Dollard, John, and Miller, N. E., Personality and psychotherapy, New Haven, 

Yale University Press, 1952. 

3. Estes, W. K., Toward a statistical theory of learning, Psychol Rev., 1950, 57, 


4. Farber, I. E., and Spence, K. W., Complex learning and conditioning as a 

function of anxiety, J. exp. Psychol, 1953, 45, 120-126. 

5. Grice, G. R., An experimental study of the gradient of reinforcement in maze 

learning, J. exp. Psychol, 1942, 30, 475-489. 

6. Gulliksen, Harold, A rational equation of the learning curve based on Thorn- 

dike's law of effect, /. gen. Psychol, 1934, II, 395-431. 

7. Guthrie, E. R., The psychology of learning, New York, Harper and Brothers, 


8. Householder, A. S., and Landahl, H. D., Mathematical biophysics of the 

central nervous system, Mathematical Biophysics Monogr., Series 1, Bloom- 
ington, Ind., The Principia Press, 1945. 

9. Hull, C. L., Principles of behavior, New York, Appleton-Century, 1943. 

Spence 21 

10. Hull, C. L., Essentials of behavior, New Haven, Yale University Press, 1950. 

11. Hull, C. L., A behavior system, New Haven, Yale University Press, 1952. 

12. Miller, Neal E., and Bollard, John, Social learning and imitation, New Haven, 

Yale University Press, 1941. 

13. Montague, E. K., The role of anxiety in serial rote learning, 7. exp. Psychol, 

1953, 45, 91-97. 

14. Perm, C. T., Behavior potentiality as a joint function of the amount of train- 

ing and the degree of hunger at the time of extinction, 7. exp. Psychol., 

1942, 30, 93-113. 

15. Pitts, W., A general theory of learning and conditioning, Psychometrika, 

1943, 8, 1-18, 131-140. 

16. Rashevsky, N., Mathematical biophysics, Chicago, University of Chicago 

Press, 1938. 

17. Spence, K. W., The postulates and methods of "behaviorism," Psychol. Rev., 

1948, 55, 67-78. 

18. Spence, K. W., Mathematical formulations of learning phenomena, Psychol. 

Rev., 1952, 59, 152-160. 

19. Spence, K. W., The nature of discrimination learning in animals, Psychol. 

Rev., 1936, 43, 427-449. 

20. Spence, K. W., and Taylor, Janet A., Anxiety and strength of the UCS as 

determiners of the amount of eyelid conditioning, 7. exp. Psychol., 1951, 
42, 183-188. 

21. Spence, K. W., and Farber, I. E., Conditioning and extinction as a function 

of anxiety, 7. exp. Psychol, 1953, 45, 116-120. 

22. Taylor, Janet A., The relationship of anxiety to the conditioned eyelid re- 

sponse, 7. exp. Psychol, 1951, 41, 81-92. 

23. Taylor, Janet A., and Spence, K. W., The relationship of anxiety level to 

performance in serial learning, 7. exp. Psychol, 1952, 44, 61-64. 

24. Thompson, M., Learning as a function of the absolute and relative amounts 

of work, 7. exp. Psychol, 1944, 34, 506-515. 

25. Thurstone, L. L., The learning function, 7. gen. Psychol, 1930, 3, 469-493. 

26. Tolman, E. C., Operational behaviorism and current trends in psychology, 

Proc. 25th Anniv. Celebration Inaug. Grad. Stud., Los Angeles, The Uni- 
versity of Southern California, 1936, pp. 89-103. 

27. Tolman, E. C., The determiners of behavior at a choice point, Psychol Rev., 

1938, 45, 1-41. 

28. Tolman, E. C., "Discussion," 7. Pers., 1949, 18, 48-50. 

Stimulus-Response Theory as 
Applied to Perception 



I should like in this paper to consider a type of perceptual activity 
which seems to be of major concern to many practicing clinical psy- 
chologists but which as yet has not been systematically handled within 
the concepts of stimulus-response psychology. I am referring to the 
fact that, given a particular complex environmental situation, the 
client will react to a certain aspect of the situation and disregard other 
aspects. Out of all the events that have occurred in some series of 
social interactions he may be alert only to those which imply some 
criticism of himself; he may even twist events, warping them so that 
they fit readily into this inaccurate perceptual schema. Given such 
selected and distorted perceptual data, he is in a sense logically justi- 
fied in reacting, let us say, with antagonism to his fellow beings. As 
a result many a clinician has concluded that the key to maladjust- 
ment lies in perception, and that if only the client could be made to 
perceive his environment in a more accurate manner, his maladjustive 
responses would be eliminated. In a sense this clinician seems to 
feel that certain kinds of responses are appropriate to certain 
kinds of perceptions, whereas other responses are appropriate to 
other perceptions; change perception and you change behavior. 

When the clinician thinks in this fashion he implicitly or explicitly 
assumes that there is no necessary and invariant connection between 
a certain physical environment and a certain perception. He is draw- 
ing a distinction between the physical environment and the behavioral 
environment [8]. Such distinctions seem necessary. 

Now it is certainly true of early stimulus-response theory that there 
was little if any deference paid to the possibility of variance between 
physical stimulus and some perceptual activity in the organism. In- 
deed, the word "perception" almost seems not to exist in the vocabu- 
lary of this early theory. To a certain degree it was admitted into* 


Wickens 23 

Hull's system [6, 7] with his differentiation of the capital S and the 
lower case s, the capital referring to the physical stimulus, the lower 
case to the resultant neural activity in the organism. Hull's concept 
of afferent neural interaction, S, goes even farther. When we recall 
that his HR, his /#, and his ER have as their prefixes the lower case s, 
we can see that at least to a limited degree this more recent formula- 
tion of S-H theory contained a perceptual term. I think Guthrie had 
always included such a term in his concept of the movement-produced 
stimuli to which responses become attached [4]. 

Even though there are these provisions for perceptual data in S-R 
theories, the research which has been done within the framework of 
these types of theory has usually disregarded perceptual problems. 
I do not mean to imply that this disregard vitiates much of the ex- 
perimental work, for actually what these experimenters have done 
for the most part is to control the factors which might lead to per- 
ceptual variance. These controls have been achieved through the 
selection of certain restricted types of situations in which to perform 
their experiments, through pretraining, and through the directions 
given to the subjects. As a result these researchers have, I think, 
been able to operate as if the stimulus were the physical event. My 
biases lead me to feel that such work has led to the identification of a 
number of important and basic principles of behavior. 

Actually my purpose in this paper is to attempt to employ some of 
these same principles to predict certain perceptual phenomena. Es- 
sentially I am taking the position that these perceptual characteristics 
with which our hypothetical clinician is concerned are molar phe- 
nomena which can be predicted from certain of the molecular S R 
postulates. Needless to say, I shall hereafter attempt to make my 
task an easier one by forgetting about the hypothetical clinician and 
client, and choose my examples from simpler experimental situations. 
They are the type of experimental situations which are described as 
offering evidence for the operation of perceptual sets or perceptual 
biases. These sets and biases are indicated in these experiments by the 
tendency of the subject to respond positively to certain characteristics of 
a complex stimulus object and to disregard other aspects of it; in other 
words, to behave in the laboratory situation as the unadjusted client 
behaves in his everyday life. Our experimental examples have a real 
advantage over the clinical cases in that the experimenter, by manipu- 
lating his experimental conditions, throws some light on the conditions 
necessary for such perceptual biases. Such was true of the studies by 

24 Kentucky Symposium 

Lawrence [9, 10] and a later experiment by Eckstrand [1] in which 
human beings rather than rats were the subjects. 

The behavior under consideration will be illustrated by an analysis 
of an experimental situation similar to the one used by Eckstrand [1]. 
Subjects are presented singly with stimuli to which they are to react 
by pressing one of three keys. For each stimulus one and only one 
key or response is correct. Actually there are nine different stimuli, 
but they are made by combining three forms with three colors. The 
solution to the problem consists of responding only to the color aspect 
of the stimulus regardless of form. The experimental design is il- 
lustrated in Table 1. We will assume that the colors employed are 

TABLE 1 Design of the stimulus-response relationships in the hypothetical experi- 
ment used to develop a perceptual bias to respond, to the dimension of color in a 

transfer situation 

Irrelevant Stiin ul i 

Relevant Correct 

Stimuli FI FZ FZ Response 

n b 

yellow, green, and blue, and that the forms are triangle, circle, and 
square. To the yellow stimulus the only response that is reinforced 
is R a , regardless of the form with which it is associated; any other 
response is not reinforced. The green stimulus is reinforced only if 
the response R f , is made, and the blue only if R (; is made. Hereafter 
these colors will be referred to by the letter C and appropriate sub- 
script, and the forms by the letter F and appropriate subscript. 

The results of the experiment by Eckstrand suggest that, after sub- 
jects have solved this problem and arc given a new problem in which 
different colors and different forms are employed, they will learn it 
with greater ease than if the first problem had not been presented 
and solved [1], One may say that these subjects have learned to 
pay attention to color, have developed a concept that color is the 
key to the solution of the problem, or are perceiving selectively. The 
question is whether it is possible to predict this behavior from molecu- 
lar S-R concepts. 

The following assumptions will be made in applying S-R theory to 
this area of behavior: 

1. If reinforcement occurs following a response to a stimulus, the 
tendency to make that response to that stimulus is strengthened. The 

Yellow triangle 
Green triangle 
Blue triangle 

Yellow eirele 
Green eirele 
Bine circle 

Yellow square 
Green square 
Bine square 

Wickens 25 

strengthening is non-selective in nature, and it occurs for those aspects 
of the total stimulus which are arbitrarily wrong as well as those 
which are arbitrarily right. 

2. Conversely, if the response is not followed by reinforcement, the 
tendency to make that response to that stimulus will decrease. 

3. Stimulus generalization will occur under both the strengthening 
and the weakening conditions. The form of the gradient I will em- 
ploy is one which is bell-shaped rather than concave upward as is 
Hovland's [5], This gradient is based upon some recent research 
conducted at Ohio State by Wickens, Schroeder, and Snide. I am 
also using the same form and shape of the generalization curve for 
excitation and inhibition. This may be incorrect. Actually the exact 
form of the generalization gradients is not important at this stage 
of the analysis, and if later research should indicate that such gradients 
do not hold generally, no major modification would need to be made. 

4. The increment in habit strength gained from one reinforcement 
is equal to the decrement resulting from one non-reinforcement. This 
assumption is not crucial, and it is made only for purposes of sim- 
plifying the exposition. Exact empirical data concerning this as- 
sumption are lacking. 

5. The learning is continuous in nature and not dependent upon 
the subject's hypothesis [11]. Indeed, the central position taken in 
this paper is that hypotheses are not the cause of the learning, but are 
in the nature of the subject's verbal expression of the habits which 
have been acquired. 


Using these assumptions, we will now consider the strength of the 
various habits the subject has acquired after twenty-seven stimulus- 
response events. We will assume that during these twenty-seven 
events each of the nine stimuli has been presented three times, and 
each of the three responses has been made once to each of the nine 
stimuli. There is nothing essential about this sequence of events, and 
the arguments presented would hold for another pattern of responses 
as well; it is arbitrarily chosen for purposes of simplifying the exposi- 

Table 2 summarizes the results of this procedure in terms of the 
occurrence of reinforcement and non-reinforcement associated with 
each stimulus-response relationship. It will be noted that reinforce- 

26 Kentucky Symposium 

ment was given for the response R a only when it was made in the 
presence of the stimulus Ci. At all other times this response occurred, 
it did not receive reinforcement. A similar state of affairs holds for 
2 and R& and for C 3 and R c . 

TABLE 2 The pattern of reinforcement and non-reinforcement associated with 
each stimulus and each response following 27 trials 



Color Form 



We will now consider what effect this sequence of events has on the 
tendency to evoke the R a response only over the entire dimension of 
color. Three reinforcements of this response were given at point Ci 






Fig. 1. 

Cj 2 C 3 

Dimension of color 

The separate inhibitory and excitatory tendencies to make the response 
R across the dimension of color. 

on the continuum, thus producing a generalization gradient of ex- 
citatory tendencies across the entire continuum. This is the gradient 
shown above the zero line in Fig. 1. However, three noil-reinforce- 



merits were given at points C L > and C 3 , and these generate the two 
curves below the zero line.* If these three gradients are summated 
algebraically, the resultant tendencies to make R a across the entire 
dimension of color are shown in Fig. 2. It is apparent that there is a 
tendency to make this response to stimuli in a range close to Ci, but 
to avoid making this response in the ranges surrounding Co and C 3 . 





Dimension of form 

Fig. 2. The algebraic sum of the inhibitory and excitatory tendencies to make the 
response R a across the dimension of color. 

A similar result will occur for C 2 and its response R b and for C 3 and 
its response R c . When these three separate S-R connections are 
placed upon the same continuum, the curves shown in Fig. 3 are 
generated. The curves indicate that there is a tendency for one or 
the other responses to be evoked by new stimuli along most of the C 
dimension. Thus, if a group of new colors were employed, these 
colors would tend to evoke this class of responses. 

It will be noted that there are gaps in the excitatory curves in vari- 
ous portions of the continuum. These result in part because of the 
relatively high frequency of wrong or unreinforced trials to correct 
or reinforced trials. As we will point out later, these gaps will or- 
dinarily be eliminated with increased training and a consequent in- 
creased proportion of correct responses. Also, these gaps would 

* It is apparent from this and other figures that we have represented the stimuli 
as being equidistant from each other along the dimensions of both form and 
color. The arguments presented in the text are not, however, restricted to such 
a state of affairs. 


Kentucky Symposium 

have been less extensive if the generalization curves for inhibition 
were steeper than those for excitation. 

Turning now to the effects of the training upon the F stimuli, we 
find that the results are quite different in nature. As shown in Table 
2 every stimulus has been given one reinforcement to every response, 
but it has also been given two non-reinforced trials for every response. 
In other words, the inhibitory tendencies are at every point more 
frequent than the excitatory tendencies at that point. The net result 


\ / 







Dimension of color 
Fig. 3. Response tendencies for all color stimuli across the color dimension. 

of this treatment is inhibition across the entire continuum, and the 
curve shown in Fig. 4 is generated. This curve will be the same for 
all of the F stimuli. 

The structure of this training has therefore developed habit tend- 
encies within the subject such that, if he is presented with a new 
problem using new colors and new forms, responses are not likely to 
be evoked by the aspect of form, but they are likely to be evoked by 
the aspect of color. This kind of behavior has the characteristics of 
a perceptual set to respond to color. It is a form of selective percep- 


As the subject begins to respond correctly in this situation, and re- 
inforcements pile up to the exclusion of non-reinforcements, a change 
will occur in the nature of the response tendencies to both the C and 
the F stimuli. 

Wickens 29 

One effect the increasing proportions of correct responses will have 
upon the tendencies to respond to the relevant or C stimulus is to de- 
crease the inhibition associated with any particular C stimulus and 
the two wrong responses. In our example considerable inhibition 
has been built up for the S-R connections of C^-R*, and Ci-R c . As 
these erroneous responses are dropped out because of the increasing 




Dimension of form 

Fig. 4. The algebraic sum of the inhibitory and excitatory tendencies to make 
any response across the dimension of form. 

strength of the Ci~R connection, less and less inhibition will accrue 
to them. 

This effect will be even more marked in the instance of the F or 
irrelevant stimuli. These stimuli ride along with the relevant ones, 
and each time a correct response is made to the relevant stimulus the 
associated irrelevant stimulus profits from the consequent reinforce- 
ment. Gradually the inhibitory potentials associated with the irrele- 
vant stimuli will be wiped out, and they will be replaced by excitatory 
potentials. The patterns of distribution of the excitatory potentials 
associated with these stimuli across the dimension will differ markedly 
from that characteristic of the relevant stimuli. The distribution for 
the relevant stimuli is asymmetrical, with particular stimuli being 
positively associated with certain responses and negatively associated 
with others. Since, however, each irrelevant stimulus is associated 
equally often with each relevant stimulus, and thereby reinforced at 
each response point, the distribution of tendencies will be symmetri- 
cal. Each F stimulus will develop an equal tendency to produce each 

30 Kentucky Symposium 

In summary, as N increases (and in the ordinary course of events 
the number of correct or reinforced responses increases), the irrele- 
vant stimuli will begin to acquire excitatory tendencies. Since, how- 
ever, particular relevant S-R connections already have a head start, 
we should not expect that subjects would eventually fall into error 
and begin responding to the irrelevant or F stimuli. There is evidence 
in an experiment by Grant and Berg [3] that increased training makes 
it easier for the subject to use as relevant a dimension that has pre- 
viously been irrelevant. 


According to the concept of response generalization, when a par- 
ticular S-R connection is reinforced, the tendency to give similar re- 
sponses to the stimulus is also strengthened. Although this concept 
suffers in usefulness because of the difficulty of rating responses on 
their degree of similarity, it has some empirical support [12, 13]. In 
the present hypothetical experiment, one would seem justified in as- 
suming that the pressing of one key is highly similar to the pressing 
of another key. If such is the case, this mechanism would, as the 
subject begins to respond correctly, serve to decrease even more the 
inhibition originally built up for particular relevant stimuli and the 
wrong responses. Thus, when to Ci the response R ff is made and 
reinforced, excitatory potential spreads to R b and R c . The magnitude 
of increment would not be as great for these two responses as for 
R , however. This mechanism would so raise the excitatory level 
that the gaps of inhibition shown between C 5 and Co and also be- 
tween C 2 and C 3 in Fig. 3 would no longer exist. Thus response 
generalization increases the tendency to make key-pressing responses 
across the entire dimension of color. It would also tend to eliminate 
some of the asymmetry of the pattern of response potential associated 
with the C stimuli. Since the F stimuli are already symmetrical, the 
mechanism would simply add a constant to the distribution of in- 
hibitory potential without changing the symmetry of the distribution. 


The magnitude of transfer from one task to a new one will be, ac- 
cording to this analysis, a function of the location of the new 
stimuli on the dimension and the responses that are required to be 

Wickens 31 

made to them [2]. In a new task a problem may be set for the sub- 
ject in which the relevant cues are drawn from the same dimension as 
those employed in the first task. These stimuli may be located at 
points CU C V9 and C z in Fig. 3. If the response R a is to be given to 
Cap, R 6 to C y , and R c to C g , then a considerable amount of positive 
transfer will occur. If, however, the responses required are not con- 
sonant with the original learning ( for example, R a to C, R& to C^, and 
R G to Cy), negative transfer or less positive transfer will occur [2]. 
This latter hedging statement arises from several considerations. One 
of these considerations is response generalization which would 
heighten the tendency to make any response to any of the relevant 
stimuli. Another is that human subjects may verbalize, saying per- 
haps, "Its color" rather than "Yellow for the middle key." If the 
verbalization is correct, this mediating response that is based on 
previous learning would be reinforced, strengthening the entire di- 
mension of color, 

In general it will be noted that this analysis does not completely 
free perception of the overt responses which have become attached to 
the various stimuli on the dimension. It implies that perceptual trans- 
fer would not be the same in experiments where the first and second 
task utilized the same responses as in experiments wherein the responses 
on the two tasks differed. Response generalization and language 
mediation would, however, lessen such differences. 


Under actual life conditions it is very likely that the complete sym- 
metry of association of the irrelevant cues with the relevant cues may 
not occur. In fact it is probable that in the usual life situation sym- 
metry of the sort that can be attained in laboratory experiments is 
the exception rather than the rule. More often than not, the natural 
situation is one in which one particular irrelevant stimulus is more 
frequently associated with one particular relevant stimulus than with 
any of the others. Thus the experimental design previously described 
might be modified so that FI was associated with Ci 60 per cent of the 
time that Ci occurred, and was associated with each of the other 
stimuli 20 per cent of the time. F 2 could be linked to C 2 , and F 3 to C 3 , 
in a similar fashion. If this were done, then each F stimulus would 
receive a greater proportion of reinforcements and a smaller propor- 
tion of non-reinforcements for one particular response than for the 

32 Kentucky Symposium 

other two. Each of the F stimuli would require relatively stronger 
tendencies to elicit one particular response than to elicit either of 
the other two responses. 

The same procedure may be employed for this type of situation as 
that described in the first section, the procedure in which each set of 
color stimuli is associated with each response an equal number of 
times. In the present case, to achieve proper counterbalancing each 
C stimulus will be associated with each response five times and will 

TABLE 3 The frequency of reinforcement or non-remforcet/icnf associated with 
each irrelevant stimulus in the 31-1 design 

Responses Made 

Presented R a R b R c 

Color Form + - + - + - 

Ci (15) Fi (9) 3 3 3 

F a (3) 1 1 1 

F 3 (3) 1 11 

C t (15) F! (3) 1 1 1 

F 2 (9) 3 3 3 

F 3 (3) 1 1 1 

C 8 (15) Fi(3) 1 1 1 

F 2 (3) 1 1 1 

F 8 (9) 3 3 3 

be reinforced or non-reinforced according to whether or not that 
response is correct for that stimulus. 

The results for the present arrangement of relevant and irrelevant 
stimuli in terms of frequency of reinforcement and non-reinforcement 
for each F stimulus are given in Table 3. In this table the first major 
column denotes the stimuli presented; this column is subdivided into 
the color and form aspects of the stimulus. The numbers in paren- 
theses indicate the frequency of presentation for each aspect. Thus 
Ci is presented a total of fifteen times, nine times with FI and three 
times each with F a and F 3 . The next group of columns designates 
what responses are made and whether or not they are reinforced. 
Note that R a is always reinforced if made in the presence of Cj, but 
responses R c and R b are never reinforced if Cj is present. 

I shall not be concerned with the relevant stimuli in this design, 
but only the irrelevant ones, and the stimulus FI in particular. Going 
down the first major column, we see that Ft receives three reinforce- 



ments in conjunction with R a (these when it is associated with Ci) 
and two non-reinforcements (these when it is associated with C 2 and 
C 3 ). The predominance of reinforcements arises, of course, from its 
linkage with the relevant stimulus C 3 . The result is greater excitatory 
than inhibitory strength for the S-R connection Fi-R a . In the second 
major column, under R h , however, there are a total of four non-rein- 
forcements to one reinforcement. The same holds true for R c . The 
resultant gradients are illustrated in Fig. 5, and they are summated 
algebraically in Fig. 6. 




Dimension of form 

Fig. 5. The separate inhibitory and excitatory tendencies to make the response 
R a across the dimension of form in the asymmetrical distribution of irrelevant cues. 

A similar treatment for F 2 and R b , and for F 3 and R c , would also 
result in this asymmetrical patterning, R& having its excitatory peak 
in conjunction with F 2 , and R c in conjunction with F 3 . It is apparent 
that the over-all curves to be drawn for these F stimuli would be 
similar to the curves in Fig. 3. They would differ, however, in that 
the magnitude of excitatory potential would not be as high as in 
the former curve. 

As the proportion of times is increased that any given irrelevant 
stimulus is associated with a particular relevant stimulus, the amount 
of excitatory potential acquired by the irrelevant stimuli increases. 
Thus, if the proportion had been 6-1-1 rather than 3-1-1 in the above 
example, the asymmetry of the habit loadings of the irrelevant stimuli 
would be even more marked and would become even more similar to 
the loadings for the relevant stimuli. 


Kentucky Symposium 

The present analysis implies that there is a basic inadequacy in 
the perceptual type of interpretation which classifies stimuli as either 
relevant or irrelevant. This S-R interpretation leads to the assump- 
tion of a continuity between the poles of the two extremes which for 
convenience will be called relevant or irrelevant. 

I should like to mention an experiment that Mr. Harold Babb is 
doing at Ohio State on the effect of varying degrees of association of 
an irrelevant stimulus with the relevant one. The experiment is 





txj *^2 ^3 

Dimension of color 

Fig. 6. The algebraic sum of the inhibitory and excitatory tendencies to make the 
response R a across the dimension of form in the asymmetrical distribution of 

irrelevant cues. 

modeled after those done by Lawrence [9, 10]. Rats are the subjects, 
and their task is to learn a discrimination. During the first part of 
the experiment the discriminative cue is the presence or absence of 
some chains hanging in the alley. For the control group the alleys 
are gray, but for three experimental groups they are black or white. 
For one of these groups white is associated with chains, the correct 
cue, 30 per cent of the time; for another, 50 per cent i.e., randomly; 
and for the third, 70 per cent. In the second part of the experiment 
the animals learn a black-white discrimination with white positive. 
The experiment is as yet incomplete, but the trend is for the 30 and 
50 per cent groups to have the greatest difficulty in learning the 
black-white discrimination, the control group to rank next, and the 
70 per cent to show the greatest ease. Thus, at least so far, the re- 
sults conform with theoretical predictions. 

Wickens 35 


There is a considerable amount of evidence which indicates that a 
one-to-one relationship does not exist between the physical character- 
istics of a complex stimulus and its stimulating value for the organism. 
Instead the stimulating value seems to depend upon perceptual char- 
acteristics of the organism. This paper is an attempt to view the 
perceptual responses as mediating responses to which the overt re- 
sponses are made, but which are predictable from a knowledge of 
the prior experiences of the organism. It is further assumed that the 
molecular postulates of S-R psychology, primarily the postulates 
dealing with reinforcement, non-reinforcement, and stimulus gen- 
eralization, may be employed in making these predictions. 


1. Eckstrand, G. A., Cue attention habits as a factor in training, A. F. Technical 

Report 6566, August 1951. 

2. Gagne, R. M., Baker, K. E., and Foster, H., On the relation between similarity 

and transfer of training in the learning of discriminative motor tasks, 
Psychol Rev., 1950, 57, 67-79. 

3. Grant, D. A., and Berg, E., A behavioral analysis of degree of reinforcement 

and ease of shifting to new responses in a Weigl-type card-sorting problem, 
J. exp. Psychol, 1948, 38, 404-411. 

4. Guthrie, E. R., The psychology of learning, New York, Harper and Brothers, 


5. Hovland, C. I., The generalization of conditioned responses: I The sensory 

generalization of conditioned responses with varying frequencies of tone, 
/. gen. Psychol, 1937, 17, 125-148. 

6. Hull, C. L,, Principles of behavior, New York, Appleton-Century, 1943. 

7. Hull, C. L., Essentials of behavior, New Haven, Yale University Press, 1951. 

8. Koffka, K., Principles of geatalt psychology, New York, Harcourt, Brace, 1935. 

9. Lawrence, D. H., Acquired distinctiveness of cues: I Transfer between dis- 

criminations on the basis of familiarity with the stimulus, J. exp. Psychol, 
1945, 39, 770-784. 

10. Lawrence, D. H., Acquired distinctiveness of cues: II Selective association in 

a constant stimulus situation, J. exp. Psychol, 1950, 40, 175-188. 

11. Spence, K. W., The nature of discrimination learning in animals, Psychol 

Rev., 1936, 43, 427-449. 

12. Underwood, B. J., and Hughes, R. H., Gradients of generalized verbal re- 

sponses, Amer. J. Psychol, 1950, 63, 442-430. 

13. Wickens, D. D., Stimulus identity as related to response specificity and re- 

sponse generalization, /. exp. Psychol, 1948, 38, 389-394. 

Motivational Forces 
Underlying Learning 


During the last few decades there has been a steady growth of interest 
in the formulation of theories designed to provide integration within 
and among limited psychological areas, including learning, motivation, 
and personality. The nature and forms of the psychological theories 
that have appeared during this time have been extremely diverse, and 
even the specific theories advanced by a single man or school have 
often changed radically within one or two decades. Although some 
psychologists may be disturbed by the evanescent trends of behavioral 
science theories, such trends are to be expected in a science as rela- 
tively young and unstructured as psychology today. 

Even when psychologists have attempted to formulate theories of 
admittedly limited scope theories primarily limited to a single area 
as learning, and to a single animal, as the rat they have not yet 
succeeded in integrating the materials into a single, orderly arrange- 
ment which even they regard as entirely satisfactory or complete. 
These limitations which psychologists face in theory construction are 
magnified when they attempt broader theories designed to relate 
diverse psychological areas and the behaviors of diverse species of 

Whatever limitations psychological theorists face and these are ad- 
mittedly many, there are advantages to be gained from attempts to 
interrelate such broad areas as learning, motivation, personality, and 
clinical research. One of the most important of the advantages lies 
in the fact that theories, even tenuous theories, often enable us to 
regard collected data from new points of view. Furthermore, new 
theoretical formulations invariably suggest additional experimental 
approaches and lead to new and different values being placed upon 
both past and proposed experimental programs. This organizing 
and evaluative function that theories inevitably have may in the long 
run harm or help any science. At its worst, theory may orient psy- 


Harlow 37 

chologists toward some line of research that fails to. uncover new, 
significant factual materials or to lead to the discovery of new prin- 
ciples which effectively organize or interrelate any broad body of 
information. At its best, theory does lead to the discovery of new 
organizing principles and new, significant facts and functional rela- 

If we attempt to integrate learning theory and motivational theory 
and orient both these broad areas toward the analysis of personality, 
we find ourselves of necessity forced to evaluate the adequacy and 
importance of the earlier psychological research in the fields of learn- 
ing and motivation to decide what lines of research diligently pur- 
sued in the past offer little hope of advancing personality theory, and 
to select neglected areas of research that offer such hope of making 
new contributions that they merit intensive exploration and exploita- 

We are particularly interested in the interrelated motivational and 
learning mechanisms having a direct role in personality formation. 
These complex problems lie within an area in which much of the 
past research has been directed toward relatively unproductive goals 
and in which potentially rich research possibilities have been almost 
totally neglected. 

If we are ever to formulate an effective theory of human personality 
and to interrelate within this area the roles of learning and motiva- 
tion, we must give proper attention and balance to all motivational 
mechanisms, and wo must certainly not ignore the motivational mech- 
anisms which are probably most intimately associated with human 
learning and the formation of human personality structure. 

Motivational theory and research have in the past put undue em- 
phasis upon the role of the homeostatic drives hunger, thirst, sex, 
and elimination as forces energizing and directing human behavior. 
There are, indeed, some psychological theorists who would have us 
believe that all or most of our adult human motives are either di- 
rectly dependent upon these homeostatic drives, or are second- or 
third-order derived drives conditioned upon visceral needs. The 
fact that derived drives based on homeostatic needs are unstable 
and transient, the fact that the conditioned drive stimulus does not 
apparently reinstate the unlearned drive state [7, 11], and the fact 
that human beings learn and live for days, weeks, or months with- 
out or in spite of a particular homeostatic need state do not disturb 
such psychological theorists in the least. It is, of course, the privilege 

38 Kentucky Symposium 

of theorists to look at man from the point of view of the pylorus or 
to look at the pylorus from the point of view of man. 

Recently, some psychological theorists, including Mowrer [9] and 
Brown [I], have emphasized another area of motivationpain and 
conditioned pain or anxiety and some psychological theorists have 
argued that these motivational mechanisms could directly or indi- 
rectly underlie much human learning and could be extremely power- 
ful forces in shaping human personality structure. No psychologist 
will underestimate the importance of conditioned pain (or anxiety, if 
it is to be so defined), because this derived drive appears to be far 
less susceptible to experimental extinction than are the derived drives 
based on the visceral need states. But lest the role of conditioned 
pain and fear be overestimated, it should be pointed out that common 
sense tells us that the greater part of our energies are motivated by 
positive goals, not escape from fear and threat. Furthermore, we 
would emphasize that intense emotional states are theoretically un- 
satisfactory motives for many learned activities, particularly learned 
activities of a moderate or high degree of complexity. It has been 
recognized for subhuman animals from the time of the formulation 
of the Yerkes-Dodson law [12] that an inverse relationship exists be- 
tween the intensity of motive and the complexity of task that can be 
efficiently learned. 

It is certainly not our desire to underestimate the importance of 
either visceral drive states or emotional conditions as motivating 
forces underlying learning and influencing personality formation. 
Visceral drive states are important motivating mechanisms in chil- 
dren, and they become important motivational mechanisms in human 
adults under deprivation. Furthermore, the appetitive mechanisms, 
innate and acquired, which are associated with (even though not of 
necessity derived from) the visceral need states are important and 
persistent human motivational mechanisms. Emotional conditions 
including pain, fear, anger, and frustration are also important and 
persistent human motivational mechanisms. 

But above and beyond the visceral need-appetitive motivational 
mechanisms and the emotional motivational mechanisms there is, we 
believe, a third major category of motives, a category of motives 
which are elicited by external stimuli and which have been described 
by such names as manipulation, exploration, curiosity, and play. 

Psychology has doubtless suffered from the fact that these motiva- 
tional mechanisms came to be labeled as "instinctive." When psy- 
chologists outlawed the term "instinct," they ruled out externally 

Harlow 39 

elicited drives as psychological motivating mechanisms and- abandoned 
research in this vital area. Prior to 1915 cultural traditions had caused 
psychologists to repress sex and talk about curiosity; after 1915 it 
became popular among psychologists to talk about sex and repress 

The denial of the existence or importance of the externally elicited 
motives is amazing because at the common-sense level of humor and 
aphorism there are many references to the operation of the external- 
drive mechanisms. It is recognized that all primates, including man, 
spend a large amount of time just "monkeying around'' and that 
monkeying around is an activity often leading to invention and cre- 
ativity. There are countless cartoons bearing on the theme that the 
monkeys in the cage stare at the people outside and are just as 
amused by what they see as are the people. Kohler reported staring 
through a peephole to see what a chimp was doing and found that it 
was staring at him! Visual exploration drives in subhuman primates 
are clearly recognized by the saying "Monkey see, monkey do." Yet, 
in spite of the obvious existence of the external drives psychologists 
have persisted in limiting themselves to endlessly repeating, with 
insignificant variations, experiments designed to show the allegedly 
overwhelming importance of the homeostatic, internal drives. But 
give psychologists their due; they have at least had enough curiosity 
to "ape" each other's work. 

Every comparative psychologist who has adapted rats for maze 
experiments knows that the rodents frequently run down the straight- 
away path used in the adaptation procedure, ignore the food, and con- 
tinue to explore the environment. Frequently, the rat will refuse to 
eat until exploration and curiosity are sated, although it may have 
been deprived of food for 23 hours previously. 

Furthermore, such behavior is by no means limited to rats; indeed, 
the prepotence of curiosity over hunger probably occurs more fre- 
quently in man than in any other animal. Those of you who have 
children know that if you deprive them of food for 3 or 4 or 14 hours 
and then seat them at the table, they will frequently engage in such 
activities as dropping their green beans into their milk, pouring their 
milk into a glass of orange juice, dangling their fork in their cup, 
throwing their spoon on the floor, and using their potatoes as a me- 
dium for finger painting. But in spite of the fact that the problem 
of dawdling at meals concerns the child psychologist, no child psy- 
chologist has ever conducted an experimental study of curiosity, 
manipulation, or exploration in the child. How psychologists expect 

40 Kentucky Symposium 

to understand human motivation and the relation between human 
motivation and learning and personality, as long as they refuse to 
study one of the most basic and important motivational mechanisms 
that the human being possesses, must and will remain a mystery 
as will also the motivational mechanism. 

Fortunately, psychologists during the last few years have shown an 
ever-increasing interest in the experimental exploration and exploita- 
tion of the area of the externally elicited drives. Schoenfeld, An- 
tonitis, and Bersh [10] have demonstrated that rats placed in a 
Skinner box will repetitively depress the bar even though this act 
produces no tangible reward. Mote and Finger [8] reported some 
years ago that rats actually decreased their running time on a straight- 
away maze from trial 1 to trial 2 even though no food rewards were 
given, and they stated that "the rats were impelled by some explora- 
tory drive to make the running response. . . ." Keller [6], Zeaman 
and House [IS], and Flynn and Jerome [3] have independently dem- 
onstrated that rats can learn when motivated only by a difference in 
illumination in the external environment. 

During the last few years we have carried out a series of studies 
on the manipulation or exploration drives in monkeys and have dem- 
onstrated that monkeys learn to solve simple and complex mechanical 
puzzles when given no reward other than the opportunity of disas- 
sembling the apparatus. A complex six-device puzzle is shown in 
Fig. 1, and a learning curve for two rhesus monkeys is given in Fig. 2. 

The performances of two groups of four monkeys on a five- 
device puzzle situation were compared by Gately [4]. One group 
was food-rewarded for puzzle solution, the other group given no 
extrinsic reward. The food-rewarded group learned more rapidly 
than the non-food group, but the best performer in the non-food 
group was the equal of any animal in the food group. Performance 
by the best member of the unrewarded group is illustrated in Figs. 
3 and 4. 

Recently we measured the performance of a chimpanzee on a 
three-device puzzle (Fig. 5). The animal was tested 20 minutes a 
day for 10 days. Five seconds after a puzzle was solved it was reset. 
Correct responses, errors, and number of puzzle completions were 
measured. The first 5 days' testing was conducted in early afternoon, 
and the last 5 days' testing between 8 and 9 P.M. Total puzzle com- 
pletions and errorless puzzle completions are contained in Fig. 6, 
which shows that the chimpanzee performed at a high level of effi- 
ciency from day 6 on, a phenomenon also illustrated by the learning 



curve of Fig. 7. The puzzle problem was probably mastered on 
day 2, for the animal made 4 errorless solutions during the last 8 of 

Fig. 1. Six-device puzzle. 







I I I I I I 





Fig. 2. Learning of six-device complex puzzle. 


the 17 puzzle completions on that day. On one day the chimpanzee 
showed constructive exploration. After solving the problem, the ani- 
mal replaced the hasp and struggled in vain to replace the hook, 

42 Kentucky Symposium 

which is, of course, an unsolvable problem when the pin is in place. 
Constructive, or anabolic, behavior, we believe, will be found to be 
highly correlated with intellectual ability, more highly in all prob- 

Fig. 3. Performance on multiple-device puzzle. 

ability than efficiency of performance on many simple learning prob- 

Harlow and McClearn [5] have demonstrated that rhesus monkeys 
can solve discrimination problems when no other incentives than 
manipulation or exploration are offered. As illustrated in Fig. 8, the 
monkey is presented with a panel holding five pairs of screw eyes, 
one member of each pair being a particular color, as red, the other a 
different color, as green. The screw eyes of the correct color are re- 



movable, and those of the incorrect color are fixed. TKree rhesus 
monkeys were tested on each of seven such sets during four 5-min- 
ute trials a day for 4 days. Figure 9 shows that significant day-to-day 
improvement in performance on problems 1 through 5 took place, and 

Fig. 4. Performance on multiple-device puzzle. 

day 1 performance on problems 6 and 7 was superior to day 1 perform- 
ance on problems 1 through 5. The mean number of responses made 
by the monkeys on the seven successive problems is plotted in Fig. 10. 
Motivation, in so far as it is measured by response frequency, is as 
high at the end of the experiment as at the beginning. 

A series of investigations of discrimination learning by rhesus 
monkeys reinforced by visual exploration is being conducted at the 
Wisconsin Laboratory at the present time by Dr. Robert Butler [2], 
The essential apparatus, a wire cage covered by an opaque box, is 

44 Kentucky Symposium 

shown in Fig. 11. The front of the box has two hinged windows which 
are covered by differentially colored cards or Cellophane. The win- 


I 26 

= 18 


E 10 
< 6 


Fig. 5. Three-device puzzle. 

I I I I I I 

I I 1 

I I 

123456789 10 11 

Fig. 6. Learning of puzzle by chimpanzee. 

dows open outward to light pressure unless blocked by a locking device. 
An opaque screen which can be raised or lowered by the experimenter 
is immediately in front of the window inside the lightproof box. Test 



trial procedure is as follows, The opaque screen is raised, and the 
monkey is given 5 minutes to respond to the colored windows. If 


I I I I I I I 

10 11 

Fig. 7. Learning of puzzle by chimpanzee. 

Fig. 8. Discrimination-learning problem. 

an incorrect response is made, the locking device is activated and a 
small light outside the apparatus is turned on. The experimenter then 
lowers the opaque screen and waits 30 seconds before beginning the 


Kentucky Symposium 

next trial. If the monkey opens the correct window, it is rewarded 
by being permitted to look out for 30 seconds. At the end of this 

Fig. 9. Discrimination learning by monkeys. 



I I I I | 1 









Fig. 10. Motivation as measured by response frequency. 

visual reward, the opaque screen is lowered, and 30 seconds later 
another trial is initiated. 

Discrimination learning in two rhesus monkeys tested 20 trials a 
day for 20 days is plotted in Fig. 12; these data show reasonably pre- 
cise and efficient discrimination learning to no other incentive than 
visual and possibly auditory search. 



The demonstration that learning to visual exploration isr possible 
led to the investigation of the strength of this motive in rhesus mon- 
keys. Two rhesus monkeys were tested by Dr. Butler 4 hours a day, 
5 days a week, utilizing the test procedures previously described. 
The measure of motivational response was log latency of response 
from the time of raising the inner opaque screen to contact with a 

Fig. 11. Visual-exploration apparatus. 

window, and Fig. 13 shows that the mean log latencies of the 5 test 
days did not increase during the day for either monkey if the total 
daily responses are divided into fourths. Performance on successive 
days' testing is depicted in Fig. 14. Motivational strength, in so far as 
it is measured by latency of response, showed an increase for one 
subject and no decrease for the other. 

Although learning was investigated only incidentally in this study, 
the data presented in Fig. 15 demonstrate rather rapid learning and 
extremely proficient performance subsequently in monkey 102. The 
other subject, 147, apparently learned, as indicated by long series of 
errorless runs, but did not maintain highly consistent performance 
over any block of 100 trials. 

48 Kentucky Symposium 

Satiation tests have been run by Dr. Butler on two monkeys thus 
far, "satiation" being defined as failure by the monkey to respond to 





I I I I I I I I I I I I I I I I A I I I I 


2 4 6 8 10 12 14 16 18 20 22 

Fig. 12. Discrimination learning to visual exploration. 



I - 8 



Rhesus 10g_ 

f 0.6 

~ *" ~ *"" ^"^^^^ ~ 


^Rhesus 147 - 



c 0.4 


1 1 1 1 

1 2 3 

Quarter periods of total trials 

Fig. 13. Visual motivation as measured by response latency. 

a window within the allotted 5-minute period on two successive trials. 
The persistence of the visual exploration motivation far exceeded pre- 



dictions. One monkey responded continuously for 9 hoursr, and the 
other for 19.5 hours, as shown in Fig. 16. 

Initial steps toward the quantification of the factors underlying 
visual exploration have recently been undertaken by Dr. Butler. The 






Fig. 14. Response latency as a function of days. 







I I \ I 
Rhesus 102 


I I 










123456 7 8 9 10 11 12 
Blocks of 100 trials 

Fig. 15. Learning to visual exploration. 

basic apparatus used in these researches differs from that illustrated 
in Fig. 11 only in the fact that there is a single, hinged door, centered 
in the face of the apparatus, rather than two doors. Facing and con- 
necting the front of the basic apparatus is a box 48 X 36 X 30 inches 
with top and sides covered by heavy black cloth and illuminated by 
a 100-watt lamp. 


Kentucky Symposium 

Each trial was initiated by raising an opaque screen which exposed 
the single window; when the monkeys pushed open the window, they 
were permitted to explore the illuminated box visually for 5 seconds. 
Four observational conditions were measured: (1) an empty box con- 
trol condition, (2) an array of five foods, (3) a moving toy electric 
train consisting of an engine and two cars, and (4) another monkey 
confined in a transport cage. 






Rhesus 167 





2 4 6 8 10 12 14 16 18 

Fig. 16. Satiation of visual-exploration drive. 


Eight rhesus monkeys were tested 30 minutes a day for 20 days. 
Each incentive condition was continued for 5 days, and order and 
frequency of incentive conditions were balanced. 

As indicated in Fig. 17, strength of incentive as measured by fre- 
quency of responses ranged downward from monkey, to train, to food, 
to empty control. Frequency of response to the monkey and train 
observational conditions was significantly greater than frequency of 
response to the empty control condition at the 0.01 per cent confi- 
dence level. Motivational strength, as measured by the total daily 
number of responses, showed no decreases throughout the experi- 

Now I am certain that every learning theoretician present has come 
to realize that the facts concerning exteroceptive motivation are in 
complete accord, or almost complete accord, with his theoretical posi- 



tion. I am certain that Dr. Spence can describe how they -fit drive- 
reduction theory, and I have tried to make this easy by speaking of 
manipulation drive and visual exploration drive, in spite of reserva- 
tions I have about the appropriateness of the term "drive" in this 
context. I am equally certain that Dr. Tolman could explain how 
our data fit within the framework of expectancy theory, and I am 
certainly certain that Dr. Guthrie could explain how they fit Guthrie 





Empty Food Train Monkey 

Visual incentive in chamber 

Fig. 17. Factors influencing strength of visual-exploration drive. 

theorybecause they really do fit the basic motivational principle of 
Guthrie theory. However, I do not believe, and do not intend to be- 
lieve, Guthrie theory no matter how well the Wisconsin data support 
the position. More explicitly, I endorse the Guthrie position that 
animals learn responses regardless of the locus of the stimulating 
agent, but beyond this my own theoretical position has little in com- 
mon with that of Guthrie. Perhaps our data will illustrate a psycho- 
logical law, and this psychological law is one that has great general- 
ity. Psychological theories seldom unify facts, but facts often unify 
psychological theories. 

Before the professional theorists explain our data to us, however, I 
should like to point out a number of facts. It appears to be most un- 
likely that exteroceptive motives can be explained as second-order, 

52 Kentucky Symposium 

derived motives conditioned to hunger, thirst, or any other primary 
drive. The manipulatory and visual exploratory drives are extremely 
persistent, whereas derived drives conditioned to so-called hunger or 
thirst show relatively rapid extinction. Indeed, one of the important 
theoretical aspects of the manipulatory and visual exploratory drives 
resides in the fact that they may show increasing motivational strength 
rather than decreasing motivational strength with repeated elicitation. 
This phenomenon would make it necessary for us to postulate some 
such phenomenon as Inverse Experimental Extinction if we are to 
account for manipulatory and visual exploration drives in terms of 
derived motives. 

We are convinced that the externally elicited motivational systems 
are as fundamental and as innate as are the hunger-appetite and 
thirst-appetite systems. By this we do not mean that learning does 
not operate as a component in the externally elicited motivational 
systems of the adult animal; we merely mean that the learning com- 
ponent here is probably no larger than it is for the hunger-appetite 
motivational system. We firmly believe that externally elicited moti- 
vational systems interact with the hunger-appetite and thirst-appetite 
systems, but we do not believe, and there is no evidence to support 
the position, that any of these systems is derived from the other, or 
that any differential degree of dependence exists among them. 

Recently, motivation theorists and personality theorists have shown 
some motivational obsession about anxiety. No one will deny that 
anxiety, or many other emotional states, serve as motivesbut any 
assumption that anxiety has some special, prepotent, motivational 
role has yet to be established. At best, anxiety is a motive for avoidant 
behavior, and the greatest part of human motivation is positive search- 
ing toward goals, not mere avoidance. In spite of our faith in the 
importance of positive, forward-oriented motives such as curiosity, 
manipulation, and exploration, we do not wish to put any constraints, 
in research or theory, upon the anxious psychologists. We merely 
wish to insist, however, that even if some psychologists are scourged 
to their experimental dungeons like quarry slaves, the remainder of 
the population will continue to be motivated by pleasant and positive 


1. Brown, C. J., Learnable drives, Nebraska Symposium, 1952. 

2. Butler, R. A., Discrimination learning by rhesus monkeys to visual-exploration 

motivation, J. comp. physiol. Psychol, 1953, 46, 95-98. 

Harlow 53 

3. Flynn, J. P., and Jerome, E. A., Learning in an automatic multiple-choice 

box with light as incentive, J. comp. physiol. Psychol, 1952, 45, 336-340. 

4. Gately, M. J., Manipulation drive in experimentally naive rhesus monkeys, 

M.A. thesis, University of Wisconsin, 1950. 

5. Harlow, H. F., and McClearn, G. E., Object discriminations learned by mon- 

keys on the basis of manipulation motives, J. comp. physiol. Psychol., 1953, 
in press. 

6. Keller, F. S., Light aversion in the white rat, Psychol Rec., 1941, 4, 235-250. 

7. Miles, R. C., and Wickens, D. D., Effect of a secondary reinforcer on the 

primary hunger drive, /. comp. physiol. Psychol., 1953, 46, 7779. 

8. Mote, F. A., and Finger, F. W., Exploratory drive and secondary reinforce- 

ment in the acquisition and extinction of a simple running response, J. exp. 
Psychol, 1942, 31, 57-69. 

9. Mowrer, O. H., Learning theory and personality dynamics, New York, Ronald 

Press, 1950. 

10. Schoenfeld, W. M., Antonitis, J. J., and Bersh, P. J., Unconditioned response 

rate of the white rat in a bar-pressing apparatus, J. comp. physiol. Psychol., 
1950, 43, 41-48. 

11. Simon, C. W., Wickens, D. D , Brown, U., and Pcnnock, L., Effect of the 

secondary reinforcing agents on the primary thirst drive, J. comp. physiol. 
Psychol, 1951, 44, 67-70. 

12. Yerkes, R. M., and Dodson, J. D., The* relation of strength of stimulus to 

rapidity of habit formation, J. comp. Neur. Psychol, 1908, 18, 459. 

13. Zeaman, D., and House, B. J., Response latency at zero drive after varying 

numbers of reinforcements, J. c\p. Psychol, 1950, 40, 570-583. 

The Premature Crystallization of 
Learning Theory 


In recent years a great deal of emphasis has been placed on develop- 
ing a learning theory, and we may actually classify certain psychol- 
ogists as learning theorists. Many of our courses in psychology now 
devote more time to discussing the relative merits of various learning 
theories than to the facts of learning, and our graduate students are 
required to develop skills in applying several theories to a given set 
of data. Thus the student's knowledge often is judged by how well 
he knows what different psychologists think and not by how well he 
is acquainted with experimental subject matter. Granting that both 
theories and experimental facts are important, what constitutes a 
healthy balance? I personally feel that an interest in theories is de- 
sirable for the development of science because theories help us 
organize facts and they help us to ask good research questions. 
However, an interest in theories can become a liability if it prevents 
us from exploring certain kinds of relationships or causes us to ignore 
facts that do not fit the theory with which we identify ourselves. 
When these things occur, the theory becomes an attitude and ideas 
become good or bad rather than right or wrong. 

Perhaps we are somewhat overambitious and have assumed that 
psychology is more advanced than facts warrant. We seem to want 
a learning theory that works not only for all learning situations but 
also for all behavior. We seem to want to predict, to do research by 
stating hypotheses, and seem no longer to be content with asking 
questions of the universe and getting our answers through research. 
When we ask questions we can be open-minded and are likely to 
let the answer to one question influence the nature of the next. It 
is in this way that we get acquainted with our universe. However, 
when we predict we show our maturity, and we can even determine 
a scientist's success by calculating his percentage of correct predic- 
tions. But we must not act more mature than we really are. 


Maier 55 

We want a learning theory to live by, despite the fact that maze 
learning, conditioning, discrimination learning, and mastering a prob- 
lem box are uncorrelated functions and despite the fact that various 
learning theorists developed their viewpoints by studying the learn- 
ing of particular animals in situations which were rather specific and 
which delimited the animals' responses. 

It would seem wiser if we would generalize somewhat less and see 
whether agreement could be reached if we did not mix data obtained 
from such a great variety of situations. Thus the fact that an animal 
shows insight in a maze does not mean it must behave insightfully in 
a puzzle box. May not random behavior and insightful behavior be 
different in kind? At least we should not make an assumption one 
way or the other until such a question has been adequately explored 
and evaluated. 

In a previous paper [6] I tried to show that there are at least five 
unrelated functions which operate in learning situations and that 
these functions are involved to different degrees in various learning 
tests. These functions are variability, plasticity, perception, behavior 
repertoire, and association fixation. (These functions are in addition 
to reasoning [7] and frustration-induced fixations [11, 12, 13] which I 
exclude from learning proper.) Since a test score measures the re- 
sultant of the various possible functions it would seem unwise to 
formulate theories which assumed the score to be a measure of a single 
function. Such an assumption discourages exploration of basic issues 
about what goes on in the animal when he solves a problem. 

To illustrate how a well-developed theory may lead to oversimpli- 
fication let us take a brief look at the treatment of reasoning given 
by Dollard and Miller [3, p. 111]. The problem has to do with a per- 
son caught in a traffic jam who needs to make a left turn onto a 
crowded highway. He notices that cars coming toward him have no 
trouble making a right turn onto the highway. The driver is caused 
to say to himself, "If I were only going the other way, it would be so 
easy!" Once this statement is made the problem is one which the 
driver has had "a great deal of practice in solving." He merely gets 
into the right-hand lane, drives past the traffic jam, turns around at 
an uncrowded intersection, comes back the other way, makes a right 
turn, and reaches his objective. 

I feel that two things have been ignored by this analysis. First, 
what causes a person to ask himself good questions? Many other re- 
sponses can be elicited by a situation of this sort. Surely a selective 
process must operate, and a selection of the right question by random 

56 Kentucky Symposium 

trial and error would be a long process. As a matter of fact, almost 
all the research on reasoning has hinged on the study of this selection 
mechanism which Dollard and Miller by-passed. Second, does the 
solution to the problem used require the highest mental process or 
might reproductive thinking be adequate? Certainly it is not the 
kind of problem used by investigators who consider the * reasoning 
process more complex than the learning process. 

Dollard and Miller's treatment of fixated behavior reveals a similar 
circumvention. In a footnote devoted to this subject they point out 
that confusion arises in interpreting the fixated behavior of my rats 
because of failure "to note that they were rewarded on every trial 
(irrespective of whether they jump to the* correct window or not) 
by escaping from the punishing air blast/' This statement assumes 
that I could not explain my rat's persistent incorrect behavior in a 
discrimination problem and formulated a non-reinforcement theory, 
without recognizing that escape from an air blast was a reward. As 
a matter of fact this point was too obvious for me to overlook. T even 
speculated over some complex probability notions suggested by Hum- 
phreys [4, 5]. However, the phenomenon of persisting behavior is 
not the fact that gave rise to my theory. The case for fixated behavior, 
right from the beginning [13], was based upon (a) the appearance of 
a different number of rats with fixated responses in the two training 
conditions used ( both of which involved the escape from the air blast 
mentioned above); and (b) the appearance of bimodal distributions 
in each of the groups. Since then additional facts have been reported, 
but these followed the initiation of the theory. 


I should like to take this opportunity to bring together some studies 
which I believe tend further to point up the need for exploring wider 
areas before jelling too specifically on a learning theory. I believe 
that these experiments bear on learning theory but would not be sug- 
gested by learning theory. In this sense 1 am making a case for 
keeping a learning theory more open-ended. One must leave room 
for empiricism and be willing to work with what Conant [1] calls 
"fuzzy" ideas. He points out that "clear-cut operational definitions 
are never possible in the infancy of a science" [p. 73] and suggests 
that they may be a handicap. 

Maier 57 

The experimental work on abnormal fixation has suggested a kind 
of behavior which does not follow learning principles as they are now 
conceived. This observation was in part based upon varied evidence 
which when combined indicated that position responses developed 
under frustrating conditions are more difficult to change than position 
responses developed under conditions of motivation or reward train- 
ing. Recently we retested this conclusion, using a large number of 
rats which matched responses but developed under the two different 

A total of 49 rats were rewarded with food for choosing the right- 
(or left-) hand card and punished for choosing the left- (or right-) 
hand card in a discrimination apparatus. After nearly 160 reinforce- 
ments the position response was firmly established and the rats were 
required to learn a card response. It was found that after 200 trials 
34.7 per cent learned the card discrimination in from 30 to 150 trials, 
49.0 per cent persisted in the position response, and 16.3 per cent 
showed variable behavior. 

A total of 55 rats developed position responses under insoluble 
problem conditions. After a similar number of trials they too were 
required to learn a card response. Only 5.5 per cent were able to 
learn the card discrimination, 92.7 per cent persisted in the position 
response, and 1.8 per cent showed variable behavior. That this dif- 
ference was not due to some difference in the learning or need condi- 
tion of the animals in the two groups was shown by the fact that 
animals failing to adopt the card-discrimination response nevertheless 
learned to discriminate between the two cards. They expressed this 
learning by jumping directly to the positive card and abortively to the 
negative card. Furthermore, the differential abortive jumping was 
shown to an equal degree by animals with and without position fixa- 
tions. Thus fixated behavior does not prevent learning from taking 
place but rather prevents certain learned behavior from becoming ex- 

We previously concluded, on the basis of this and other evidence, 
that fixated behavior is of a compulsive nature. If one assumes that 
frustration-induced fixations have a compulsive property it follows 
that learning which does not violate the compulsive tendency will be 
expressed more readily than learning which violates it. 

Mr. Paul Ellen is exploring this possibility. After an animal has 
developed a persistent position response and has failed to adopt a 
card-discrimination response, but has expressed its discrimination by 
jumping abortively to the negative card and directly to the positive 

58 Kentucky Symposium 

card, it is confronted with a three-window situation. Suppose that a 
particular animal always jumps to the right window of a pair when 
the jumping stand is directly in front of the two windows. (This jump 
to the right occurs even when the positive card is on the left and the 
negative card is on the right.) If we now confront such an animal 
with three windows and place the jumping stand directly in* front of 
the first two, then there will be two windows to the animal's right. If 
the first, second, and third windows contain positive, negative, and 
positive cards, respectively, it is possible for the rat to make a jump 
to the right and to the positive card at the same time (by jumping 
to the extreme right). By using various combinations of three win- 
dows in this manner Mr. Ellen was able to cause a fixated rat to fol- 
low the positive card as it was moved between the second and third 
positions. Eventually the fixation was broken when the rat was able 
to follow the positive card to the left. Thus by training the animal 
to respond within the bounds of its fixation it became possible to break 
some fixations at least. 

It seems that the whole problem of persistent behavior must be 
carefully explored with a more empirical approach. There are per- 
haps many kinds of rigidity,* and attempts to account for compulsive 
behavior in terms of our present learning concepts may make them 
unacceptable to clinicians. 

In recent years I have made much of the point that frustrating or 
insoluble problem situations divide a group of individuals [11 9 12], 

* Possible mechanisms giving rise to persisting or rigid behavior are the fol- 

(a) Relatively greater strength of response in question over other responses 
in organism's repertoire. 

( b ) No motivation to change is present in situation. 

(c) No other responses in organism's repertoire. 

( d) Motivation to persist is present in the form of fear of change. 

(e) Compulsiveness created by frustration. 

The postulation of a variety of persistent behaviors also assumes basic differ- 
ences in procedures for correcting them. These corrective procedures may 
take the following forms: 

(a) Increasing the habit strength of alternatives by repeated reinforcement. 

(b) Altering the animal's needs (change from hunger to thirst), which should 
cause shift in behavior even if situation is unchanged. 

(c) Exposing animal to additional learning or knowledge (e.g., exploration of 
new territory). 

(d) Removal of factors giving rise to fear (e.g., a permissive environment). 

(e) Removing state of frustration, guidance, and other procedures under in- 

Maier 59 

be they rats or college students, into two populations: (a) those 
whose subsequent learning scores fall within the range of a normal 
distribution curve, and (b) those whose scores fall outside the dis- 
tribution curve and show them to be clearly handicapped in their 
learning. We explain this break in a group by saying that a given 
situation frustrates some individuals and not others. Thus when an 
individual's frustration threshold is exceeded he moves from one 
distribution curve to another. Marquart [14] previously reported 
that a period of random punishment ( electric shocks ) on 75 per cent 
of the trials in an insoluble problem disturbed the subsequent learn- 
ing scores of college students when the problem was made soluble. 
Seventy-four per cent of the students performed like a control or an 
unpunished group, but the other 26 per cent formed a new popula- 
tion and made an average score more than 4 times that of the control 

Recently Marquart and Arnold [15] repeated this experiment with 
another group of students, but instead of electric shock punishment 
they used a light signal to indicate an incorrect choice. Again a 
bimodal distribution of scores was obtained when learning on a sol- 
uble problem was tested. The number of individuals falling near 
one mode or the other was related to the extent of failure experienced 
by three different groups, but the point at which there was a gap in 
score between the two modes was the same. In this instance the 
proportion of normal to extreme performers was 77 to 23 per cent, or 
almost exactly the same as in the first study. Thus a period of failure, 
not punishment as such, causes about one-fourth of a group of persons 
to behave qualitatively differently from the other three-quarters. 

That the presence or absence of frustration is a determining factor 
in whether or not a given individual performs normally or exceedingly 
poorly was suggested both by the responses of the individuals during 
the experiment and by the relationship of extreme scores with certain 
personality variables. Certainly an attempt to explain the results in 
terms of differences in learning aptitude is unrealistic. 

In another experiment with animals, Ellen and I tested the relative 
influence on learning of variations in reinforcement and of variations 
in the opportunity to perceive a change in effect. Animals were 
trained to form a position response. After 160 trials an attempt was 
made to replace this response with a card-discrimination response. 
Three training methods were selected which seemed adequate to 
evaluate the reinforcement and perceptual functions. 

60 Kentucky Symposium 

Method A. The positive card placed on the side of the previously 
learned position response on 8 out of the 10 trials given daily and the 
negative card on the position side for the other 2 trials. This method 
results in a reward '.punishment ratio of 80:20 as long as the animal 
continues to express a position response. 

Method B. The positive card placed on the side of the position re- 
sponse in 5 out of each 10 trials and the negative card on the position 
side for the other 5 trials. This method results in a reward '.punish- 
ment ratio of 50:50. 

Method C. The positive card placed 011 the side of the position re- 
sponse in 2 out of each 10 trials and the negative card on the position 
side on 8 trials. This method results in a re ward '.punishment ratio 
of 20:80. 

These three methods differ in the degree of reinforcement given the 
position response as well as in the absolute number of times that a 
card response is reinforced. Thus, regardless of how one wishes to 
treat reinforcement, including partial reinforcement [./#], one should 
expect animals to be able to learn the new response as well as give 
up the old one. Further, Methods A and C should be more different 
from each other than either is from Method B. 

If, however, the perception of inconsistency is important to learn- 
ing, then Methods A and C should be more like each other than either 
is like Method B. By Methods A and C the same thing happens on 
nearly all trials, either almost always reward for a position response 
or almost always punishment for a position response. Thus the world 
is fairly orderly and consistent, which means that the animal's expecta- 
tions are nearly always fulfilled. The most inconsistent condition is 
Method B y when either effect (reward or punishment) is just as 
probable as the other. Should this irregularity inspire learning, in 
that the need for learning is great, or might it be discouraging? 

The results obtained are shown in Table 1. First, it is clear that 
none of the conditions assured learning, although there was more than 
ample time for this. Thus a given training procedure teaches some 
animals and entirely fails with others. Such wide differences in learn- 
ing performance cannot be set aside by assuming that a bimodal 
distribution of aptitudes exists. 

Second, since the results of Method B do not lie between those of 
Methods A and C, it is apparent that there is no trend which follows 
the degree of reinforcement. Rather, Methods A and C are strikingly 
similar with regard to both the number of animals abandoning the 

Maier 61 

position response and the number adopting the card response." If the 
results of Methods A and C are combined they are significantly dif- 
ferent from those of Method B. (P = 1 per cent for abandoning the 
position response and between 1 and 2 per cent for adopting the new 
response.) Thus it seems that the method yielding the most incon- 
sistent effect caused the fewest number of animals to change. 

One might suppose that this failure to cause so many animals to 
learn is due to the fact that Method B is more difficult than the other 
two. However, the learning data (last column) exclude this possi- 
bility. Not only are there no significant differences in learning scores, 

TABLE I EJfcul of 1/irw reward-pimishrrutnt ratios on behavwr modification 





Trials to 

R :1> Ratio 






















but also the trend suggests that condition B makes for faster learning. 
Method B thus seems both to facilitate learning and to disturb learn- 
ing more than the other methods, and which effect it produces appears 
to be a function of the individual. It is for this reason that insightful 
problem solving, trial and error, and frustration-instigated behavior 
cannot be described in situational terms. Any of these behaviors can 
be stimulated by a situation in which there is thwarting or blocking 
of goal-oriented behavior [10, 11]. What occurs in a given individual 
depends on such factors as (a) physiological differences, which may 
merely be threshold differences in emotion; (b) personality differ- 
ences, which may reflect different degrees of determination; (c) in- 
tellectual differences, which may determine the relative difficulty of 
the problem for the individual; (d) perceptual differences, which 
may be determined not only by past experiences but also by chance 
mental sets and minor neurological differences; (e) need differences, 
which vary from individual to individual as well as with the period 
of deprivation; and (/) the repertoire of learned responses. Because 
all these processes influence what an individual learns or how he 
changes as a consequence of an exposure to a situation they are rele- 
vant to learning, yet only the last two are generally given full recog- 

62 Kentucky Symposium 

In previous studies from our laboratory [2, 6, 8, 9, 17] reference has 
been made to perception as a determiner of what is learned. Experi- 
mental data [8, 9, 17] previously published showed that various rats 
in mastering the same discrimination problem actually came away 
with different learning, as tested by the method of equivalent stimuli.* 
The same studies also showed that the difficulty of a discrimination 
problem, as well as what was learned, differed, depending upon which 
of a pair of stimuli was made positive. 

It seems easier to handle these variables and relationships in a per- 
ception theory than in a learning theory, and I feel that present learn- 
ing theory discourages research which investigates these matters. 
However, today I wish to go even a step farther in complicating mat- 
ters for learning theory and indicate how the selection or expression 
of behavior can be influenced by an emotional tone. 

Zucker [19] has demonstrated that delinquent children completed 
some of his stories differently from non-delinquent children. William 
Edwards, one of our graduate students, gave Zucker's completion 
stories to three groups of sixth-grade children (mean age 12.6 years) 
in a role-playing situation. One group was instructed to feel rejected 
by their parents, another group was instructed to feel wanted or ac- 
cepted by their parents, and the third group was not instructed. 

The results obtained are presented in Table 2 along with Zucker's 
data. From this table one can see that the story-completion results 
for the "rejected" group were significantly different from those for the 
"accepted" group, whereas the uninstructed group made scores part 
way between those of the other two groups but somewhat closer to 
the scores of the "rejected" group. 

Further, the results of the "rejected" group were strikingly similar 
to those of Zucker's delinquent group, and the "accepted" group com- 
pleted stories similar to those of Zucker's non-delinquent group. 
When the data for the four stories are combined we find that, on the 
average, 76 per cent of the "rejected" group gave delinquent re- 
sponses, which compares closely with Zucker's 75 per cent for actual 
delinquents. Only 32 per cent of the "accepted" group gave delin- 
quent responses, which is slightly higher than the 24 per cent ob- 
tained from Zucker's non-delinquents. 

* Differentiation between the stimulus objects could be made in terms of the 
absolute properties of either (or both) the positive or negative stimulus, relative 
over-all brightness, figure size, ground properties, etc., and the relative use made 
of these differences varied greatly from individual to individual. 



Edwards' population was chosen from a public school in which the 
deliquency rate was high. This .perhaps explains why the unin- 
structed group (control) approached the group who were instructed 
as "rejected" and why the "rejected'* group duplicated Zucker's re- 
sults with a group of actual delinquents somewhat more accurately 
than the "accepted" group duplicated his non-delinquent group. 

It seems that this study, as well as some other role-playing results 
[16], indicates that the expression of behavior can be controlled by 
manipulating attitudes or feeling states. This is indicated by the dif- 


Results of role playing , delinquency and non-delinquency 

Role-Play Data 


Xucker Data 
per cent 

per cent 





Story I 

Went away 
Kept knives 



Went to friend 















per cent 


ference in the story-completion results of the two instructed groups. 
More important, however, is the observation that Edwards' children 
seemed to have both delinquent and non-delinquent behavior in their 
repertoires. Ordinarily we speak of training delinquents to be good 
citizens and assume that children learn delinquency. Without intro- 
ducing reinforcement and without giving them experience with rele- 
vant behaviors, different responses were elicited from two groups of 
children when only the role was changed. 

Surprising too is the fact that these children "seemed" to know 
that delinquency and parental rejection went together. Where did 
they learn this? Might it not be better to say that when they felt re- 
jected they perceived the world as hostile and unfriendly and that 
the behavior elicited by these perceptions is known as delinquent be- 
havior? I personally feel that feelings of frustration produce be- 
haviors which do not have to be learned. However, they can be 
learned and they can be modified by learning, but learning is not a 
basic essential for a good deal of the behavior which is expressed. 

64 Kentucky Symposium 

To assign learning a greater role in behavior than it actually plays is 
to exclude consideration and exploration for other variables. 


Theory formulation is not necessarily a sign of progress, and it 
may actually be a disservice to science if it becomes an end in 
itself. The type of research reported in this paper suggests that the 
concept of reinforcement, so important to present learning theory, 
requires re-examination. Any attempt to incorporate association for- 
mation, motivation, and perception into a single quantitative theory 
seems premature, since our knowledge of each of these processes still 
is in a state of development. To combine them all into a reinforce- 
ment concept buries the problems rather than stimulates analysis. 

Progress in science can be made by developing principles and 
doctrines to describe qualitative relationships, and formulating laws 
to describe quantitative relationships. Such objectives of research 
permit the exploration of ideas and encourage speculation and the 
testing of hypotheses. They permit one to think of learning in con- 
nection with the counseling process and group-discussion processes 
as well as in connection with training experiments. Laws and doc- 
trines may be amended, modified, extended, or refuted with relative 
ease since these changes do not threaten a whole system. Not only 
are the fruits of such research stimulating to future research, but also 
in the meantime they may serve practical uses in education, character 
development, and therapy. 

The history of science has been one of gradual clarification of ideas. 
Theories of limited scope make their appearance, and eventually 
some of them may incorporate others or give rise to a larger theory. 
Psychology is at the stage where theories of perception, of motiva- 
tion, of association formation, of organization, of reasoning, and of 
frustration can be formulated so as to organize large bodies of knowl- 
edge. However, discrepancies and disagreements will be numerous. 

A general behavior theory is premature at the present time because 
it (a) discourages exploratory research; (fo) emphasizes quantitative 
measurement at the expense of qualitative analysis; (c) assumes that 
science develops along deductive logical lines, thereby excluding many 
other sources of development and kinds of thinking; and (d) is un- 
willing to entertain concepts which are still vague and in the process 
of development. 

Maier 65 


1. Conant, J. B., Modern science and modern man, New York, Columbia Uni- 

versity Press, 1952. 

2. Eglash, A., Perception, association, and reasoning in animal fixations, Psychol. 

Rev., 1951, 58, 424-434. 

3. Dollard, J., and Miller, N. E., Personality and psychotherapy, New York, 

McGraw-Hill, 1950. 

4. Humphreys, L. G., The effect of random alternation of reinforcement on the 

acquisition and extinction of conditioned eyelid reactions, /. exp. Psychol., 
1939, 25, 141-158. 

5. Humphreys, L. G., Acquisition and extinction of verbal expectations in a 

situation analogous to conditioning, J. exp. Psychol., 1939, 25, 294-301. 

6. Maier, N. R. F., The specific processes constituting the learning function, 

Psychol. Rev., 1939, 46, 241-252. 

7. Maier, N. R. F., The behavior mechanisms concerned with problem solving, 

Psychol Rev., 1940, 47, 43-58. 

8. Maier, N. R. F., Qualitative differences in the learning of rats in a discrimi- 

nating situation, J. comp. Psychol., 1939, 27, 289-331. 

9. Maier, N. R. F., The effect of cortical injury on equivalence reactions in rats, 

J. comp. Psychol, 1941, 32, 165-189. 

10. Maier, N. R. F., Reasoning in humans: III The mechanisms of equivalent 

stimuli and of reasoning, J. exp. Psychol, 1945, 35, 349-360. 

11. Maier, N. R. F., Frustration: the study of behavior without a goal, New York, 

McGraw-Hill, 1949. 

12. Maier, N. R. F., and Ellen, Paul, Can the anxiety-reduction theory explain 

abnormal fixation? Psychol Rev., 1951, 58, 43,5-445. 

13. Maier, N. R. F., Glaser, N. M., and Klec, J. B., Studies of abnormal be- 

havior in the rat: III The development of behavior fixations through frus- 
tration, /. exp. Psychol, 1940, 26, 521-546. 

14. Marquart, D. I., The pattern of punishment and its relation to abnormal 

fixation in adult human subjects, /. gen. Psychol, 1948, 39, 107-144. 

15. Marquart, D. L, and Arnold, L. P., A study in the frustration of human 

adults, J. gen. Psychol, 1952, 47, 43-63. 

16. Solem, A. R., The influence of the discussion leader's attitude on the out- 

come of group decision conferences, Doctoral dissertation, University of 
Michigan, 3952. 

17. Wapner, S., The differential effects of cortical injury and retesting on equiva- 

lence reactions in the rat, Psychol Monogr., 1944, 57. 

18. Wilcoxon, H. C., "Abnormal fixation" and learning, J. exp. Psychol, 1952, 

44, 324-333. 

19. Zucker, H., The emotional attachment of children to their parents as related 

to behavior and delinquency, /. Psychol, 1943, 15, 31-40. 

Learning and 


A sentiment or a psychical system or a superego is just as physical a 
notion as a cell assembly, an engram, a reflex arc, or even an atom. If 
this proposition strikes you as self-evident or even banal, so much the 
better for psychology. If it does not, at least I have something to 
talk about. 

A generation or two ago it might have been supposed that "physi- 
cal" meant "material," but in a period when Einstein's equation 
E = MC 2 appears in mass-circulation slick-paper magazines, this is 
rarely what is meant. When mass is energy and energy may be a 
local deformation of space, when the particleis it the pi-meson? 
that holds the atomic nucleus together, and thus gives "matter" such 
permanence as it displays, has itself a duration of a tiny fraction of a 
millionth of a second, the concept of materiality ceases to be physi- 
cally meaningful. However, let us call this Meaning 1. 

Consider what a psychologist means by "physical" when he talks 
about a physical (sometimes, to be sure, a physiological) model or a 
physical correlate. The word seems always, except for an occasional 
atavistic naivety, intended to mean something like "comprehensible 
to the physical principles and theory current at the time of writing." 
Actually, it is more likely to mean "comprehensible to the writer's re- 
membered high school or college physics and haphazard impressions 
of developments since then." In either event it does not mean the 
same thing on 13 March 1953 that it meant on 9 August 1935 or on 21 
November 1912. And we may be quite sure that it will mean some- 
thing still different by this time next year. If this is the meaning of 
"physical" in such expressions, we are not saying anything very defi- 
nite or precise when we use them. Let us call this Meaning 2. 

There must be several psychologists here who remember when it 
was generally regarded as a complete and final refutation of Thorn- 
dike's law of effect to remark that it required a retroactive effect of a 


Adams 67 

later state of affairs upon an earlier and that this was "physically" 
impossible. Now that the phenomenon of feedback and the construc- 
tion of servomechanisms have become physical commonplaces, we 
get along quite comfortably with the notion of reinforcement, as it is 
now fashionable to call the law of effect. It is even possible now to 
talk about the self-regulation of organisms and of subsidiary systems 
thereof without becoming liable to the horrid suspicion of vitalism- 
provided we call this concept homeostasis.* 

We are not yet in psychology far enough from an implied belief in 
primitive word magic for it to be wholly funny. It is still too stultify- 
ing to psychological theory. When we persuade ourselves that it 
makes us more objective or physiological to call a meaningful object 
a stimulus, even though it may be what it is for the behaving animal 
by virtue of a long history of varying kinds of commerce with it, we 
are indulging in what Harry M. Johnson [9] once appropriately called 
"the method of equivocance." When Pavlov [14] called a tissue need 
or a remote long-term goal a "reflex" he created a confusion that psy- 
chology has not yet won clear of. Much as we properly deplore the 
inherent antinomies and the excesses of certain semantic doctrines, 
the general injunction to keep our terms clean and their referents un- 
ambiguous can be ignored only at the price of talking nonsense. 

Let us, at all events, try to get some clarity into the meaning of 
"physical" as used in psychological discussion. Sometimes the word 
used is "physiological" or "neurological," but these are clearly re- 
garded as merely special cases or departments of physical. We are 
told, for example, that unless our hypothetical constructs are in 
neurological terms they are "forever removed from empirical investi- 
gation" [11, p. 284]. It may be useful in this connection to consider 
briefly a series of papers by David Krech, whose egregious and de- 
plorable apostasy [II, 12] from the insights he earlier attained is 
enough to make strong men weep.f 

* Why, by the way, do you suppose that Kohlcr's admirable 1920 exposition 
of self-regulation and steady states in physical systems and the eligibility of this 
concept for psychological use did not suffice to make it respectable for psycholo- 
gists whereas Cannon's later adoption of it did? Was it merely a happier choice 
of a name for the phenomenon? Or just that Cannon was duly labeled a physi- 

t Manifestly I cannot quote extensively enough to supply complete contexts 
or to be wholly fair. I can only ask Krech to forgive this for the good of our 
common cause. His reflections are so far in advance of most current psychologi- 
cal thinking that it must seem ungrateful to pick on him. But they are so far 
from his own high standard that he deserves the worst that compression im- 

68 Kentucky Symposium 

The general thesis of the second article is that "the moment we in- 
troduce hypothetical constructs into our theory building, then the 
purely psychological approach becomes untenable" [11 9 p. 287]. The 
clinching argument for Krech, several times repeated, is his question, 
"Where do these hypothetical constructs exist?" [11 9 p. 284] One is 
tempted to answer, "In the same place as the binominal theorem or 
the kinetic theory of gas." The only way I can make sense of this 
question is to suppose that it means, "Where in space are the data 
(objects, structures, processes, events, etc.) into which these con- 
structs may be expected, with advancing inquiry, to evolve?" But 
even this comes very close to implying Meaning 1 of "physical/' The 
application of the seventeenth century criterion of materiality, simple 
location, is a special case of what Whitehead has called [16, p. 85] the 
fallacy of misplaced concreteness. 

To make a long story short, the answer to Krech's crushing question 
is, "In exactly the same 'physical* world or Nature as the atom or elec- 
tron." If one has a seventeenth century model of this Nature, or a 
seventeenth century meaning for "physical," this is bound to be 
opaque. If, in my answer, "physical" is taken in Meaning 2, it is 
equivalent to believing that physics is done, finished, exhausted, that 
there will be no more Plancks, Rutherfords, or Einsteins. Somehow 
this seems unlikely. We have been hearing a good deal lately about 
binary calculators, servomechanisms, transistors, and such like. For 
all we know, or for all physiology or neurology can tell us, our brains 
or, more generally, our bodies may be and probably are full of tran- 
sistors, perhaps one in every nerve cell, and a thousand other gadgets 
that physics has not yet invented, or even dreamt of. When one 
looks back over the developments of the last 50 years of physics and 
considers that the rate of discovery appears to be increasing logarith- 
mically, one wonders why we are so timid about our constructs or 
should insist that they be reducible to the physics of today or even of 
50 years hence. Our constructs may be good physics or bad physics, 
but that does not depend on their intelligibility in terms of present- 
day physics. It depends rather on their adequacy to the data of be- 
havior all the data, all the phenomena, that we have at a given time. 

poses upon us. And his dangerous eloquence could easily corrupt the youth of 

There is much, especially in the second paper [12], with which one must 
agree. As a matter of fact, I think his current "dynamic systems'* and his earlier 
"cognitive structures" are one and the same thing, namely sentiments, of which 
more later. 

Adams 69 

Any constructs that require us to close our eyes to any of. the phe- 
nomena of experience are bad constructs and bad physics. I am 
perfectly willing to "accept the Universe" as Krech [11, p. 288] de- 
mands, but it does not seem reasonable to reject most of it for the 
sake of the pip-squeak fraction for which we have at the moment a 
partial understanding. 

It will probably be evident by now that I have smuggled in a third 
meaning of "physical," one that is neither so patently grotesque as 
Meaning 1 (materiality) nor so mercurial and stultifying as Meaning 
2, with its dependence upon the calendar. In this third sense, 
"physical" simply means "explanatory." Its opposite is "phenomenal." 
This meaning was first forced upon me in reading Kohler's The place 
of value in a world of facts, which 1 had great difficulty in under- 
standing until I read "explanatory" for "physical" in most of its occur- 
rences. Now this Meaning 3 will carry us quite a long way, perhaps 
as far as we shall get today. At least it may be sufficient to establish 
the validity of the opening statement of this paper. Its difficulty is 
the "bifurcation of Nature" that Whitehead [15] has indicated one 
kind of escape from; its implication that cognition is not a natural 
process but something that is outside of and beyond Nature and 
mediates the latter to something else called Mind with a capital M.* 
This bifurcation is another item of what Krech has somewhere called 
"Western folklore," in this case Cartesian, in which it has "split asun- 
der what nature had put together" [10, p. 82]. 

When Krech speaks of "a proper respect to present neurological 
knowledge and theory" [12, pp. 345 f.] I think he is dead wrong, in 
spite of his comprehensive and important qualifications. The only 
things to which any inquiry owes respect are its phenomena. The 
attitude of respect on the part of an empirical science is never appro- 
priate toward existing principles of its own or any other field of in- 
quiry. You break out of the bonds of a doctrine and enlarge it only 
by not having respect for it. We are inherently conservative enough 
without submitting to such restrictions. Thus Planck "five years after 
Einstein's first publication on the photon theory of light, angrily 
commented that all the fruits of Maxwell's great work would be lost 
by accepting a quantization of energy in the wavefront 'for the sake 
of a few still rather dubious speculations'" [8, p. 99]. So far as I 
know the reconciliation of the wave and the photon theories of light 
is not yet complete, and you still have to think of light in some con- 

* This must not be supposed to depreciate the utility of the really indis- 
pensable construct of "a mind" or "a personality." 

70 Kentucky Symposium 

texts as undulatory and in others as corpuscular. But no one doubts 
that the integration will somehow be effected. It is an opportunity 
for some future Maxwell. The point is that you always lose when 
you ignore phenomena for the sake of principles and may even come 
out with some such anomaly as a "miniature system," a pedagogical 
device sometimes useful in teaching mathematics but a self-contra- 
diction in empirical science [2], The phenomena that Krech seems 
in danger of ignoring in his retreat from "cognitive structures" to 
"dynamic systems" arc, of course, those of cognition, of which more 

This is all the more puzzling since later in his second paper [72] he 
welcomes and quotes Bertalanffy's suggestion that physics will en- 
rich and fructify its theorizing through attention to the phenomena 
of biology. Why does he stop there? Why not recognize (a) that 
physics must ultimately comprehend psychological phenomena and 
(b) that this must come about through the psychologizing of physics 
rather than through closing our eyes to psychological phenomena? Is 
this another manifestation of Western folklore, of the Cartesian bifur- 
cation of Nature? 

Human cognition is certainly the most highly developed, as well as 
our most accessible, paradigm for a kind of relation among natural 
objects. But the fact that physics has not yet got round to dealing 
with this relation hardly justifies us in ruling it out of Nature or, in- 
deed, in closing our eyes to something rather like it in white rats. I 
imagine one reason that the term "reinforcement" has so widely re- 
placed "law of effect" is that it permits its users more comfortably to 
evade this relation. But think how many cognitions are involved in a 
single reinforcement arid try to imagine its occurring without them. 

One more quotation from Krech and my consideration of his recent 
publications will close. He argues that the purely psychological ap- 

... is untenable because it makes forever impossible any attempt to ap- 
proach the study of our hypothetical constructs in any more direct manner 
than through the examination of the original stimulus-response correlations. 
This is so, I must repeat, because the psychological position places hypo- 
thetical constructs in a domain which, by definition [italics his], is forever 
removed from any direct observation (for that domain, it will be remem- 
bered, is neither behavioral, experiential or neurological). The conclu- 
sion seems inescapable that a psychological field or a life-space cannot 
itself be a construct nor can it provide us with the substrate in which we 
can profitably place our hypothetical constructs [11, p. 288]. 

Adams 71 

Part of this crushing dictum I have already dealt with: the psycho- 
logical field is physical in the only meaning of "physical" we have 
yet found that will make sense. There is a fourth meaning natural 
that also makes sense if you know what you mean by the concept of 
nature, and in this sense also psychological constructs are just as 
physical as neurological ones. But we should not overlook the pos- 
sibility that psychological fields may also be the only fields in nature 
that are phenomenal. Perhaps psychological field is the only con- 
struct in physics that also is directly experienced. 

What concerns me here is an epistemological fallacy that has had a 
paralyzing effect upon American psychological thinking ever since 
Watson popularized it. This is the widespread belief that our per- 
ceptions of "physical" situations and objects are somehow in better 
epistemological status than our perceptions of other people's psycho- 
logical situations and objects; that seeing a cat's situation as frustrat- 
ing is less secure in epistemological principle than seeing his physical 
environment as made of wood or glass. This is simply not so. The 
only test we have in either case is intersubjectivity. This test may be 
less frequently or less easily satisfied for psychological situations (al- 
though even this is questionable), but when it is satisfied they are 
just as objective and just as eligible to be causes as are "physical" ones. 
For psychologists especially it is important to realize that the cpiste- 
mology of values is neither different nor separable from that of things. 
There is not time to develop this notion here and now. Perhaps its 
validity is self-illuminating. 

But what does Kreeh mean by "direct observation" in the last 
quotation? Does he think an atom or an ion or a gene has been 
"directly" observed? He may have seen a streak in a Wilson cloud 
chamber that someone told him was the track of a particle of water 
after impact of an electron. lie may have seen a photograph of an 
image in an electron miscroscope that someone told him was the 
atomic lattice of a crystal. But these are a long way from "direct" 
observation. We are apt to forget what an elaborate chain of infer- 
ence is built into our instruments. And if you are going to worry 
about transmission theories of light and sound, as a Cartesian will, 
you must also worry about the transmission of primary qualities. 
When is any observation direct? And when, above all, are we going 
to quit wondering whether a given experience is possible and deal 
with the experience? 

I shall not discuss Krech's critique of the hypothesis of psycho- 
physical isomorphism because that can be done more appropriately 

72 Kentucky Symposium 

and effectively by others. My understanding of it differs somewhat 
from his. I conceive it as primarily a heuristic device, useful in the 
present state of knowledge, to generate or suggest hypotheses about 
cortical function which may then be experimentally tested by their 
implications for perception. Thus, to put it in terms that it is one of 
the aims of this discussion to transcend, its function is tot make not 
psychology but neurology, and it is of no use to psychologists qua 
psychologists, but only in so far as they turn themselves into neurolo- 
gists. Actually and historically, of course, it has had extremely valu- 
able psychological by-products, has led to important discoveries that 
would otherwise have waited a long time. 

But I have been much too hard on Krech. Actually, we are in- 
debted to him for creating a climate, providing a basis for communi- 
cation, that obviates the necessity for this discussion to be even 
more long-winded than it is. He appears to have been seduced jointly 
by MacCorquodale and MeehFs "On a distinction between hypotheti- 
cal constructs and intervening variables" [J3] and by Hebb's [7] 
persuasive exploitation of some recent developments in physical un- 
derstanding. Let us now take a brief look at these two works. 

Perhaps critical notice should have been taken of the MacCorquo- 
dale and Meehl article when it first appeared. But who could have 
foreseen that it would mislead a psychologist of Krech's demonstrated 
quality? And life is much too short, as Wallach once remarked in 
another context, to spend it rushing about trying to keep all the people 
who are so inclined from jumping off bridges. 

I believe that most of my critique of Krech's position applies with 
equal or even greater force to that of MacCorquodale and Meehl. In 
general, their point is that we must give tht? name of "construct" only 
to such hypothetical schemes as might "conceivably" be true. This 
seems reasonable enough until one realizes that their range of con- 
ceivability is much narrower even than Krech's. Notions that make 
no pretensions to represent what they call "objective existence" should 
be called "intervening variables." These authors are righteously in- 
dignant about notions, such as libido, that are introduced as innocent 
intervening variables and then subtly transformed into constructs by 
the surreptitious addition of dynamic properties. "These hypotheti- 
cal constructs, unlike intervening variables are inadmissible * because 

* Note the stern and final damnation implied by that word "inadmissible." 
That really puts them in their place. It might be wondered if the subject of 
this symposium, personality, is really admissible. 

Adams 73 

they require the existence of entities and the occurrence of processes 
which cannot be seriously believed because of other knowledge" 
[13, p. 106]. 

Don't these people realize that this inadmissible transformation of 
intervening variables into constructs is precisely the way science is 
made? A little perspective would help here. At what point did 
Gregor Mendel's intervening variable of alternative hereditary char- 
acters become a hypothetical construct? When it was baptized a 
gone? When reduction division was first observed? Or not until 
after the electron microscope? 

Structural formulae of compounds were at first written as a device 
to insure that all the "valence bonds" of the elements involved were 
used. It was really inadmissible later to regard this metaphor as 
representing the actual structure of molecules. Fortunately the 
chemists had not read MacCorquodale and Meehl and didn't know 

It would be interesting, if time permitted us to trace the fascinating 
development of the notion of the atom, to inquire at what point and 
by virtue of what inadmissible imputations of properties it became a 
construct and, indeed, beyond that, how it became a natural object.* 
Some of you may have seen textbooks printed in this century in which 
the atom was represented or even pictured as a little round ball with 
one to four hooks representing its valences, which had to be inad- 
missible notion "satisfied" for the atom to be in a stable condition. 
The serial modifications of this wretched, metaphorical intervening 
variable at the hands of Aston, Rutherford, Bohr, and their numerous 
successors are familiar to you. They have further specified the nature 
of the hooks, but the hooks are still there. 

What I think such a survey would reveal would be something like 
this: when a notion shows a good deal of versatility and seems to be 
applicable to a variely of phenomena beyond that for which it was 
designed, it becomes a valued construct, irrespective of the immedi- 
ate plausibility of the mechanisms it envisages. 

At the very end of the MacCorquodale and Meehl article, just before 
the summary which reiterates its doctrine of sterility, there is a para- 

* I called up a physicist friend the other day and asked him, "What is the 
subject matter of physics?" His response was, "The natural relations among 
inanimate objects." "And are atoms, electrons, mesons, etc., objects within 
the meaning you intend?" "Oh, sure!" he said. No epistemological misgivings 

74 Kentucky Symposium 

graph of great wisdom. Unfortunately, both the authors and Krech 
appear to have missed its significance.* Here is the paragraph: 

Of course this Judgment in itself involves a "best guess" about the future. 
A hypothetical construct which seems inherently metaphorical may in- 
volve a set of properties to which hitherto undiscovered characteristics of 
the nervous system correspond. So long as the propositions about the con- 
struct are not stated in the terms of the next lower discipline, it is always 
a possibility that the purely formal or relational content of the construct 
will find an isomorphism in such characteristics. For scientific theories 
this is enough, since here, as in physics, the associated mechanical imagery 
of the theorist is irrelevant. The tentative rejection of libido would then 
be based upon the belief that no neural process is likely to have the com- 
bination of formal properties required. Strictly speaking, this is always 
problematic when the basic science is incomplete [13, p. 106]. 

It ought to be remarked too that Krech's practice of theory construc- 
tion is far better than his preaching of it. Thus he boldly imputes 
properties (segregation, which he calls localization, rigidity, etc.) to 
his construct of dynamic systems for which little plausible basis exists 
in even the most speculative contemporary neurology, on the excellent 
ground that, plausible mechanism or no, they have to be thus and so 
to account for the phenomena; just as Mendel's alternative characters 
had to be in the germ cell to account for his data, whether or not re- 
duction division was known to contemporary cytology. Krech may 
be right or wrong about any of these imputations, but that will be 
settled by their ability to cope with the psychological data, not by 
how easily they slip into the strait jacket of contemporary neurology. 

But I must also pay my respects to Hebb, our Faust's real Meph- 
istopheles, whose winsome style, wide-ranging scholarship, acute crit- 
icism, and candid facing up to the difficulties of his doctrine are an 
almost overpowering combination. Almost, but not quite. What 
Hebb has done, very inadequately and roughly described, is to apply 
some relatively recent notions from physics, notably that of rever- 
berating circuits, to the nervous system and to test the fit for a num- 
ber of psychological phenomena. Now this sort of thing might be 
thought to come under the head of good clean fun for psychologists 
that cannot possibly do any harm and may actually advance neurol- 

* They credit Dr. Herbert Fcigl with clarification of this point. It is a curious 
and somewhat depressing commentary on contemporary psychological thinking 
that two empirical scientists have to have their puristic zeal abated and qualified 
by a methodologist. But it is very pleasant to be able to say something nice 
about a logical positivist! 

Adams 75 

ogy.* But all too frequently, in American psychology at least, the 
use of the momentarily favored neurological concept comes to be 
regarded as equivalent to the propositions: (a) psychological events 
not understandable in these terms cannot happen, and (&) psycho- 
logical constructs that cannot be put in these terms cannot be enter- 
tained, i.e., are "inadmissible." This happened in the case of the 
reflex arc, of Sherrington's idea of the final common path, of the all-or- 
nothing law. It will almost inevitably happen in the case of Hebb's 
cell assemblies, and what appeared to some as an extension of the 
possibilities of psychological theory construction seems to be already 
becoming another strait jacket, t 

But enough of methodology. It is a sterile and ineffectual business, 
and I hope that 1 have said my last word about it. It is emasculating 
too many of our bright young meii.j The men who make science 
don't seem to know much about methodology. Newlon, for example, 
said ho didn't make hypotheses. It is much more profitable to work- 
ers in a young field like psychology to read the classics, the original 
reports of discoveries in older empirical fields, than to study treatises 
on scientific method, written by sophisticated contemporary logicians 
with benefit of hindsight. We have to recognize that not only science 
but also the scientific method itself is always in the making. The 
hardest task of the discoverer is always to free himself from the 
preconceptions he is not aware of having, just because they are part 
of the intellectual climate of his time, just because they are also taken 
for granted by the treatises of scientific method. Besides, as Conant 
has remarked [5, p. 48], a propos the extraordinary tenacity of the 
phlogiston theory, "We can put it down as one of the principles 

* Psychologists can have a great deal of this kind of fun making neurological 
applications of the successive physical discoveries oi the next 50 years. 

f Even if Jlebb and Krech are making good neurology, as I suspect they are, 
in the same sense that the little hooks on the atom were good physics, the rejec- 
tion of needed psychological constructs because they do not jibe even with their 
new and adventurous neurology is still a strait jacket. Why, for example, this 
fixation on the nervous system? There are a lot of other systems in our bodies, 
about a few of which we already know a little something. Doubtless the nervous 
system is some sort of bottleneck in the determination of behavior, but it is 
hardly the whole story. 

t Frank Lloyd Wright is credited with the remark that the American culture 
is the only one that has gone from barbarism to decadence without passing 
through civilization. This, of course, is merely a libelous witticism as regards 
the culture as a whole, but it seems fairly to describe the transition of American 
psychology from Watsonian illiteracies to the contemporary preoccupation witli 

76 Kentucky Symposium 

learned from the history of science that a theory is only overthrown 
by a better theory, never merely by contradictory facts/' * If meth- 
odology really played the role its exponents ascribe to it, one con- 
tradicted implication of a theory would suffice to dispose of it. But 
we are all aware of contemporary learning theories that show the 
same kind of resistance to this supposedly fatal defect- that the 
phlogiston theory did in Priestley's time. 

So let us turn to what I hope I may, with the permission of Mac- 
Corquodale and Meehl, call the constructive part of this paper. I 
have used so much of my time for the critical part that this will have 
to consist of a series of rather bald assertions on which you arc in- 
vited to do your own reflecting. 

We (and the other animals, at least) live in a world not of stimuli 
but of objects. I would like to leave it at that; but, because of our 
unfortunate Cartesian linguistic habits, I have to add that these ob- 
jects are what they are for a given creature by virtue of the structure 
of his nervous system and sense organs, the condition of his endocrine 
system, strength, height, skin color, mood, past commerce with 
similar objects, and many other things. So I make the temporary 
concession of calling them psychological objects, and their totality at 
a given time, the animal's psychological environment or field. This 
totality, over longer periods, Lewin has called, following v. Uexkiill, 
the life-space. 

All proposed psychological laws appear to be of the form B = 
/(P, E) or, as some prefer to write it, R = /(O, S), and it is widely 
believed that all three variables must be physical in Meaning 1 or 
Meaning 2. But no such law has ever, to my knowledge, been ob- 
tained. Probably the most nearly successful such effort is Crozier's 

Lnii 55 

= fcjlog sin* a 

hm 20 

for the orientation of young rats on an inclined plane in the dark. 
The catch in this is the constant fc, which is different for different 

* It is too bad that we do not have in America the counterpart of Ostwald's 
Klassiker der exakten Wissenschaften. These arc small paper-bound reprints, 
translated into German when the original language is another, of some hun- 
dreds of the real landmarks of science, such as Galileo's Dialogs, Claude Ber- 
nard's Experimental medicine, and Faraday's, Maxwell's, and Gibbs's articles. 
They are sold for about fifteen cents and are obtainable almost anywhere. There 
might well be some connection between this fact and the great fertility of 
German science in fundamental ideas, which Bronfenbrenner [4] has noted in 
respect of psychology. 

Adams 77 

species of rat, but which is not independently determined and simply 
has the value required to make the equation fit the theoretical curve. 
The guess is hazarded that it may represent an unknown complex of 
anatomical characteristics, such as center of gravity, length of femur, 
and angle of pelvic girdle. The point is that by this device the physi- 
cal (Meaning 1 or 2) environment, angle a, is surreptitiously trans- 
formed into the psychological environment, or the environment as it 
is for the behaving animal. 

No. In order to work, psychological laws have to use psychological 
variables: i.e., acts rather than responses; organisms as personalities 
rather than as proton -electron aggregates, pieces of protoplasm, or 
cell assemblies; and objects rather than stimuli. We have to write the 
general form of behavioral laws as B^ = f(P^ 9 EJ. It may happen 
that B^ sometimes coincides with V<f> and is partially describable in 
physical (Meaning 1 or 2) terms, as in the orientation of Crozier's 
rats, where it can be partially described by angle 0. I say partially, 
because what the rat is doing is maintaining its balance; and describ- 
ing this by the aspect that happens to interest the experimenter is like 
describing the behavior of a dog trailing a fox by saying that he is 
going north or is moving his legs.* 

Now the totality of conditions in an organism that make a given 
object what it is for the behaving animal, i.e., the general case of 
Crozier's fc, I shall call a sentiment. I have been accustomed sys- 
tematically to define a sentiment as a part of a personality identified 
by its reference to an object f and to define personality by pointing, 
the most fundamental of all operations. 

Now this is a concept that, under one name or another, has been 
found indispensable by nearly all students of personality and by most 
students of social psychology. It has been variously called disposi- 
tion, mental system, psychical system, means-end readiness, attitude, 
belief, derivation, metanerg. Krech used to call it cognitive structure 
but now regards the notion as illegitimate. He has imputed some of 
its properties to his present dynamic systems, and, knowing how re- 

* Zencr has pointed out bow thinking about the conditioned response is 
vitiated by restricting consideration to one aspect of the animal's behavior [17], 

t This includes, of course, the corresponding action system: a thing is for us 
what we do with or about it. There is a very good idea at the center of the 
old motor theory of meaning. All it wants is a little cleaning up and moderation 
of its claim to be the whole story. 

It also includes the affective component, an aspect of which I have recently 
dealt with [3]: i.e., an object is also what we feel about it, as well as what we 
know and do about it. 

78 Kentucky Symposium 

sponsible he is to the phenomena, I feel safe in predicting that he 
will soon impute the rest of them. 

Now if the concept of sentiment should prove useful in the under- 
standing of learning it would have demonstrated some of the versatil- 
ity that I suggested earlier is more important than ease of translation 
into contemporary physical (Meaning 2), physiological, or neuro- 
logical terms, and, incidentally, it would further the synthesis with 
which this symposium is concerned. 

I believe that the same assumptions that are made for purposes of 
personality theory and perception suffice to further somewhat our 
understanding of learning. 

It would seem natural to suppose that my sentiment for (attitude toward, 
cognitive structure for) this building is in some sense a part of my sentiment 
for the class buildings. To put it more generally, if object A is a member 
of object B, sentiment A' is a member of (or part of, or region of) sentiment 
B f . Moreover, if object A is a property (or a sign) of object B, it seems 
evident that sentiment A' must in some sense be a member of sentiment B'. 
Thus, my sentiment for the height of this room is a part of my sentiment 
for this room. In general, it would seem that all the constitutive relations 
obtaining among a personality's objects can be ordered to relations of mem- 
bership among the sentiments corresponding to those objects. 

Of peculiar importance among these is the relation of instrumentality: if 
A is instrumental to B, positively or negatively (i.e., as means or as barrier), 
sentiment A' is a member of sentiment B'. If we use the sign - to signify 
"is a means to" and the sign d to mean "is a member of," the notion could 
be expressed by the equivalence: 

A _> it _> {} ~> j) -+ E - = ,r e B' c r e ir <= K 1 - - 

Thus, a given hierarchy of means-end relations among the objects in a per- 
son's psychological environment is represented in the personality by a senti- 
ment with precisely the same number of hierarchically organized member 
sentiments on the same number of membership levels. If all the objects of 
a given person's environment are instrumentally related to one, then all his 
sentiments are members of the sentiment for that object and the personality 
is said to be integrated [3]. 

Consider the naive dog in the classical Pavlov experiment. He is 
alone in a strange and silent room. For a more or less protracted 
period nothing happens except the events of his own body. Then sud- 
denly there is a buzzing noise off to his right. He has a sentiment 
for noises. He pricks up his ears and looks in that direction. His 
back hair may rise. Before he has done reacting to this event, an- 
other occurs: a biscuit rattles down a tube and appears in a pan im- 
mediately in front of him. He has a sentiment for biscuits and knows 
how to deal with them. 

Adams 79 

Now look at the same dog after ten or twenty such sequences of 
events. Now, when the buzzer goes off, he does not look in that di- 
rection but lowers his nose to the pan and gets the biscuit before it 
has stopped rolling. The buzzer is now another object than it was 
at first: it is a sign or property of the biscuit in much the same way 
as the taste, smell, or hardness, which had also at some time to be 
learned. And according to our assumption about instrumentality the 
dog's buzzer sentiment has been incorporated in his biscuit sentiment. 
How such irrelevant things as a buzzer and a biscuit can be inte- 
grated into one object I have described elsewhere [I, pp. 170 ff.]. 

This change in the relations or "meaning" of an object seems to be 
the one objective thing that is common to all the processes that have 
been called learning. Baptizing a concept docs not ordinarily accom- 
plish much, but occasionally it does reveal an identity that might 
otherwise escape notice. If, as I believe, the conditions within an 
organism that make an object what it is for that organism the general 
case of Crozier's k can be identified with the concept of sentiment 
that personality theory finds indispensable, a solid basis exists for the 
integration of learning theory and personality theory.* 


I. Adams, D. K., A restatement of the problem of learning, Brit. J. Psychol., 

193], 22, 150-178. 
2 Adams, D. K., Note on method, Psychol Rev., 1937, 44, 212-218. 

3. Adams, D. K., The organs of perception: sentiments, 7. Pers., 1953, 22, 5259. 

4. Bronfciibrenner, U., Toward an integrated theory of personality, in Blake, 

R. JR.., and Ramsey, G. V. (Ed.), Perception: an approach to personality, 
New York, Ronald Press, 1951, pp. 206-257. 

5. Conant, J. H., On understanding science, New York, New American Library, 


6. Crozier, W. J., The study of living organisms, in Murchison, C. (Ed.), Tlie 

foundations of experimental psychology, Worcester, Clark University Press, 
1929, 45-127. 

7. ITebb, D. O., The organization of behavior, New York, John Wiley, 1949. 

8. Holton, G., On the duality and growth of physical science, Amer. Scientist, 

1953, 43, 89-99. 

9. Johnson, H. M., Some fallacies underlying the use of psychological "tests," 

Psychol. Rev., 1928, 35, 328-337. 

* Professor Krech, who was good enough to read this paper soon after the Sym- 
posium, thinks my selection of quotations seriously misrepresents his views. Since 
he may be right and since the articles in question are exceedingly valuable in any 
case, I would urge the interested student and theorist to read them more carefully 
than Krech thinks I have. 

80 Kentucky Symposium 

10. Krech, D., Notes toward a psychological theory, J. Pers., 1949, 18, 66-87. 

11. Krech, D., Dynamic systems, psychological fields, and hypothetical constructs, 

Psychol Rev., 1950, 57, 283-290. 

12. Krech, D., Dynamic systems as open neurological systems, Paychol. Rev., 

1950, 57, 345-361. 

13. MacCorquodale, K., and Meehl, P. E., On a distinction between hypothetical 

constructs and intervening variables, Psychol. Rev., 1948, 55, 95-107. 

14. Pavlov, I., Conditioned reflexes, Oxford, 1927. 

15. Whitehead, A. N., The concept of nature, Cambridge, 1920. 

16. Whitehead, A. N., Science and the modern world, New York, Macmillan, 


17. Zener, K. E. f The significance of behavior accompanying conditioned salivary 

secretion for theories ol the conditioned response, Amcr. J. Psychol., 1937, 
50, 384-403. 

Ego Psychology, 
Cybernetics, and 
Learning Theory 

O. H. MOWHEll 

It will best serve our present purposes if I discuss the three topics 
constituting the title of this paper in the reverse order of that in 
which they are here mentioned. But first a more general word. It 
can hardly escape even commonplace observation that we tend to 
take, as models for interpreting the complex and mysterious, phe- 
nomena which are simpler and more fully understood. Hence, the 
machine, being man-made and intelligible, has often patterned our 
thinking about the less intelligible aspects of man himself. 

Today we face a new challenge in this respect. Stan ton and Sylva 
Cohn, writing in a current issue of The Scientific Monthly, put the 
matter well when they say: 

The nineteenth century was the "Age of Power." It saw the develop- 
ment of the machine, and concomitant with it there arose a mechanistic 
philosophy of life and a mechanical interpretation of life processes. . . . 

Science has advanced beyond the mechanistic stage, however. Just as 
the nineteenth century was the Age of Power, the twentieth century is the 
Age of Communication and Control. It is not enough to make a powerful 
machine, having the ability to do many times the work of man. There 
must be an intelligent application of this energy it must be controlled [1, 
p. 87]. 

In recent decades engineering has moved rapidly forward along 
these lines, producing, oddly enough, machines that are more "in- 
telligent" in practice than living organisms are in theory! Of course, 
some of these machines are actually more "intelligent," as regards 
certain specialized tasks, than are animals, including men. Here we 
think particularly of the "giant computers/' for example. But we are 
presently concerned rather with the extent to which living creatures 
are, per hypothesis, more limited in their potentialities than we know 


82 Kentucky Symposium 

them, in fact, to be. Modern machines are thus presenting a chal- 
lenge to our theorizing in psychology and related sciences. If we 
can meet this challenge, psychology may, as Harry Harlow hopefully 
opined a few years ago, "eventually catch up with common sense." 
It is with this challenge and some of the new vistas it opens up that 
this paper will be mainly concerned. 


In the paper already cited, Cohn and Cohn say: 

If there is one law that marks this era [the nineteenth century Age oi 
Power] definitively, it is the principle oi the conservation of energy. This 
principle, which is expressed in the first law of thermodynamics, has been 
characterized as the greatest generalization in natural science. But it is not 
the final word [p. 87]. 

"Habit" is a concept born of this tradition. A stimulus, as cause, 
impinges upon a sense organ (internal or external) and sets up neural 
impulses which, by virtue of certain neural "connections," travel to and 
activate certain muscles. The resulting response is the effect. Energy, 
though transmitted and transformed, is thus strictly "conserved," in 
the manner of a moving billiard ball striking and imparting its mo- 
mentum to a second ball, it to a third, and so on. But thus far we have 
not distinguished between "habit" and "reflex." Reflex, we are told, 
is invariable, umnodifiable; habit, on the contrary, can be changed. 
But how? Thorndike noted that, with a habit, the cause-effect se- 
quence does not end with response. Responses, he observed, may 
in turn initiate causal sequences in the external world which termi- 
nally impinge back upon the organism. These "feedback" effects 
Thorndike, like the layman, called rewards if they lessen stimulation 
in some important way, and punishments if they significantly increase 
stimulation. Rewards, Thorndike conjectured, strengthen the S-R 
sequences that produce them, whereas punishments have the reverse 

That an organism that can be thus modified by experience that 
can, in other words, learn will on the average be better off than a 
purely reflex organism is pretty obvious. But the model or image 
which Thorndike gave us has not been a universally satisfying one. 
On the one hand it has been charged, rather unjustifiably it seems, 
with being "teleological" (in the opprobrious sense of the term); it 
has also, more relevantly, been accused of making organisms more 

Mowrer 83 

"mechanical," "blind," "stupid" than they really are. Although cer- 
tainly more "intelligent" than a purely reflex creature, Thorndike's 
"habit" animal is by no means overburdened with brightness. Yet 
it has not been too easy to say exactly what is wrong with such a 
creature and how it might be improved. 

Here we will attempt a concise diagnosis. It follows from what 
has been said that if a Thorndikian animal has acquired, under one 
set of conditions, a given "habit" and if conditions are now changed, 
the animal itself can begin to change only after it has performed the 
old response under the new conditions at least once. In other words, 
this much stupidity, or "mal adaptation," is logically demanded by 
the theory, and has been rationalized by the slogan: "Organisms learn 
only by doing!" Evidence recently reviewed in another paper [7] 
indicates that this inference is plainly not valid and calls for a radical 
revision of what we have previously called learning theory. 

In order to escape from the Thorndikian dilemma, we must, first 
of all, repudiate one of his major assumptions: we must abandon the 
idea that rewards strengthen stimulus-response bonds and that 
punishments weaken them. What they do instead is to produce, by 
conditioning, secondary reinforcements and secondary motivations, 
respectively. This will at first hardly seem like a clarifying statement. 
What it means, quite simply, is that we do not learn, or fixate, overt, 
behavioral responses at all. These are always "subject to change," 
depending upon the "situation." What is learned are attitudes, mean- 
ings, or expectations which consist of token decrements in emotional 
tension (secondary reinforcements, or rewards) and token increments 
(secondary motivation, or punishment). It is assumed that it is these 
inner, conscious factors which, moment by moment, select and shape 
overt action; and if we take this position we have ample provision 
for "learning" without doing, i.e., for changes in behavior that occur, 
solely and immediately, because the situation, or, more exactly, the 
individual's internal tension state, or "field," has changed. Here we 
have the capacity for foresight, insight, and a generally higher order 
of intelligence and adaptivity than is possible in a "creature of habit." 

But we have achieved this at a cost which some will be reluctant 
to pay. Instead of channeling stimulus energies directly through the 
nervous system, switchboard fashion, and out into motor organs in 
a highly determined way, we are here assuming that this kind of de- 
terminism holds, so to say, only half way. It holds, I assume, to this 
extent, that meanings and attitudes, both positive (tension reducing) 
and negative (tension inducing), follow quite automatically, quite 

84 Kentucky Symposium 

reflexly (conditioned reflexly), upon the occurrence of significant 
stimuli or situations. But here this type of fixed, cause-effect relation- 
ship ends and a more complex mechanism takes over. I am sure that 
you will not hold me accountable for explaining all the riddles of 
consciousness merely because I refer to the phenomenon; but I will 
venture the guess that consciousness is, essentially, a continuous-com- 
puting device or process. The eternal question is, "What to do? How 
to act?" And consciousness, as I conceive it, is the operation whereby 
information is continuously received, evaluated, and summarized in 
the form of "decisions," "choices," "intentions" which then emerge as 
behavior. Life asks the questions, sets the problems, and it is the bus- 
iness of consciousness to give us the "answers." * 

This is not to say that consciousness is merely chance or caprice. 
In the paper already cited [7], an attempt is made to state some of 
the principles of conscious functioning, and more attention will be 
given to this problem in the later sections of the present paper. 1 
hope that I have succeeded, thus far, in showing the general direction 
in which learning theory must, in my judgment, move if our concep- 
tual models in psychology are to be as resourceful and sensible as 
the "real thing" or even as the newer types of machines that can today 
do such remarkable things. 


It will, I trust, be evident how naturally the foregoing analysis artic- 
ulates with some of the basic concepts of cybernetics. One of these 
concepts, as Wiener [10] and others have formulated it, is that only 
the simpler machines and response systems operate on a "blind," re- 
flexive principle; "higher" behavior and machines involve the "feed- 
back" principle. Learning, I propose, is not a matter of strengthening 
or weakening connections between drives and overt behavior, but of 

* The reintroduction of the concept of consciousness into behavior theory may 
seem, to some, like a retrogressive step. To such persons it may connote a return 
to introspection as the chief mode of psychological inquiry and a relinquislunent 
of the gains of half a century of intensive experimental inquiry. What I wish to 
suggest, without being able to develop at this time, is the notion that conscious- 
ness is a phenomenon which would be inferable from, and indeed logically de- 
manded by, the empirical facts, even though there were no direct experiential 
access to it whatever. The point is that adherence to a strict S-H psychology 
creates a number of dilemmas which only the interpolation of an integrative 
mechanism of some sort can resolve. Consciousness is here conceived as that 

Mowrer 85 

the acquisition of "positive" (rewarding) and "negative" (punishing) 
feedback from stimuli that have accompanied past action or experi- 
ence. If, in the past, a given act has been predominantly rewarding, 
then incidental stimuli, both external and internal, which have been 
associated with the act-reward sequence will take on, as already 
noted, the capacity to produce secondary reward and thus to guide 
or direct the organism into the same or similar action in the future. 
If, on the other hand, a given act has been predominantly punishing, 
the incidental stimuli which have been associated with the act-punish- 
ment sequence will take on the capacity to produce a secondary-drive 
increment ("fear") and will tend to guide the organism into different 
behavior.* Behavior, in any given situation, is the organism's best 
effort, then and there, to find that line of action with the greatest 
positive and the least negative feedback and hence with the best 
likelihood, when it is carried through to completion, of being maxi- 
mally satisfying and minimally hurtful. Feedback has been called 
"the central theme of cybernetics." We must, it appears, likewise 
give it a major role in psychological theory. 

Another way of saying much the same thing is to stress the role 
of communication and control. These are the factors which are con- 
spicuously missing in most nineteenth century and earlier machines 
and which are also minimal or absent in reflex action in living organ- 
isms. Central to the communication process, surely, is the phe- 
nomenon of meaning [8], and this we have given a central place in 
our theoretical scheme. Meanings, not means; attitudes, not actions 
these are the most immediate outcomes of learning; and it is 
through these that intelligent control of behavior then becomes pos- 
sible. It is not yet certain whether so-called "information theory" 
as it has been elaborated by Shannon and Weaver [9] and others will 
prove as helpful in behavior analysis as many psychologists now 
hope; but it seems abundantly clear that the more rudimentary tenets 
of cybernetics have already given our thinking in the realm of learn- 
ing theory a useful nudge; and I shall try to show in the next and 
last section of this paper that the developments thus produced carry 
us appreciably closer, as learning theorists, to ego psychology than 
we have been before. 

First, however, it will be salutary to clear up a terminological point. 
In the preceding pages I have spoken of positive and of negative feed- 
backs, defining them as secondary reward and secondary punishment, 

* These principles are here stated in highly condensed form. For elaboration 
and exemplification, the reader is referred to the longer paper already cited [7]. 

86 Kentucky Symposium 

respectively. Here it should be said that for Wiener and other cyber- 
neticists these terms have a different meaning. If a control system, 
like a thermostat or the governor on a steam engine, has a stabilizing 
influence on some quality heat, in the one case; speed, in the other- 
then the feedback is said to be negative, regardless of whether it is 
acting at any given moment to bring a quantity tip to or down to a 
standard state. The "error," in other words, which the control system 
is trying to "correct" may be either "plus" or "minus," but the type 
of over-all control is, in either case, said to involve negative feedback. 

The opposite, or positive, type of feedback is no less interesting. 
The products of certain chemical reactions are such as to have a 
catalytic effect upon and thus to speed up the reactions which pro- 
duce them. Sometimes these reactions gain momentum so rapidly 
as to produce "explosions" they are, in fact, the same in principle as 
the chain reactions in the atom bomb and the H-bomb. If, likewise, 
certain other types of processes start to slow down, the effects of the 
retardation accelerate the slowing down. Thus, if a business concern 
is not prospering, inability to pay employees satisfactory wages or to 
meet creditors may accelerate the failure. 

Since homeostasis, or self-regulation, is one of the essential char- 
acteristics of living organisms, it will be immediately evident that 
many of their activities will have the properties of negative feedback, 
as just defined. These activities or functions serve to hold certain 
states or qualities within fairly narrow limits of variation. Positive 
feedback, when the concept is applied to living organisms, sounds 
as if it would be pathological, to say the least, and, in the extreme 
case, lethal. Actually, we find some very instructive instances of it 
in living organisms. We have already seen that there is a general 
tendency for incidental stimuli which have been associated with pri- 
mary-drive reduction to produce secondary-drive reduction and for 
such stimuli as have been associated with primary-drive induction to 
produce secondary-drive induction. These two types of contiguity 
learning, or conditioning, unquestionably tend to produce (indirectly, 
through the integrating function we call consciousness) responses 
which have biological utility notably, flight or immobility in the face 
of danger and approach to objects or situations with rewarding po- 

However, there is one crucial respect in which this general scheme 
is inadequate, non-biological. The world's good things do not always 
remain conveniently at rest, waiting to be claimed and consumed. 
Sometimes these goal objects, especially when they are other organ- 

Mowrer 87 

isms, have a way, at the critical moment, of eluding their pursuers. 
Therefore, an organism that became more and more confident and 
relaxed as it approached a quarry might find itself slowing up at just 
the point when a final "push" was needed for success. It is there- 
fore interesting and altogether understandable that we should find 
the phenomenon of appetite. Its outstanding feature is an increase 
in secondary motivation just as consummation is imminent, thus giv- 
ing to behavior at this crucial moment a peculiar urgency and "oomph," 
Here, it seems, is an instance of positive feedback: "The nearer you 
get, the more yon want it" carries the proper connotation hero. Cer- 
tainly such an arrangement is biologically intelligible, but it is none- 
theless enigmatic. Neal Miller [3] has put the matter this way. We 
have reason for thinking that stimuli associated with consummatory 
states take on contradictory capacities: a tendency to cause a decre- 
ment in secondary drive (secondary reinforcement) and a tendency 
to cause an increment in secondary drive (which we call, not punish- 
ment, but appetite). As Miller points out, if these two tendencies 
occurred simultaneously, they would be self-canceling, mutually neu- 
tralizing; so his proposal is that they may alternate, producing inter- 
mittent bubbles of pleasurable anticipation and surges of intensified 
drive. This is frankly a speculation and, even if true, leaves many 
unanswered questions. However, consideration of the problem in 
the cybernetics setting illuminates and sharpens it in ways which we 
shall further explore in the next section. 


It is at once evident that neither reflexology nor habit theory, in 
the mode of Thorndike, Pavlov, Watson, or Hull, can provide a very 
sophisticated approach to ogo psychology, so-called. The reflex and 
habit, almost by definition, exclude consciousness, which is the core 
of ego functioning. The same difficulty does not at all occur in the 
conceptual framework presented in the preceding sections of this 
paper. Here we posit a division of labor between learning as a purely 
unconscious, automatic process, on the one hand, and conscious judg- 
ing, deciding, and acting, on the other. Our conception of learning 
therefore blends naturally enough with ego theory. 

Thinking, in this frame of reference, we see as a foim of activity 
in which the individual makes symbolic responses ( Hull's "pure stim- 
ulus acts") for the purpose of eliciting, or "sampling," the feedback, 
or effects, that could be expected if the action thus symbolized or rep- 

88 Kentucky Symposium 

resented were really carried out. In simplest, most primitive form, 
thinking, or reasoning, can be seen, quite literally, in the vicarious 
trial-and-error of a rat at a maze choice point a form of "light" activ- 
ity which is manifestly a prelude to grosser and more consequential 
action. In adult human beings the process becomes both more subtle 
and more elaborated, but its function, basically and ultimately, re- 
mains the same. 

Recently I heard a college president say that the most important 
thing his institution could do for students was to "teach them to 
think." At one level this can be regarded merely as an old pedagogic 
bromide. Or it can be seen as profoundly and perennial ly true. If in 
thinking we are, so to say. sticking our "mental necks" out into the 
future and "feeling around," nothing, at least on occasion, could be 
more useful, especially when the findings are later translated back 
into intelligent action. Living organisms swim forward in a sea of 
time, and those with the best "distance receptors," i.e., with the best 
symbolic skills, will almost certainly have an edge in the struggle 
lor existence. "Use your head" (or head end) is good advice both in 
the sense of using the special senses and in the sense of moving back 
and forth through time in the way that symbols make dramatically 

Then, too, our conceptual scheme puts us in a good position for 
understanding not only thought but also fantasy. The difference, 
basically, is that thought is a preparation for action; fantasy is a sub- 
stitute for it. In fantasy we select not those symbols that will forecast 
reality but rather those that will yield, without reference to reality, 
the greatest present pleasure, or secondary reinforcement. "In day- 
dreams I often picture myself as a very generous and kindly person," 
a young woman undergoing psychotherapy recently remarked. This 
was said in the context of revealing just the reverse real characteristics 
and shows the highly autistic nature of such activity.* 

So far, however, we do not yet have a psychology of the abnormal, 
a psychopathology. Does our system, as presently conceived, yield 
one? The attempts of reflexologists and habit theorists to "explain" 
neurosis and propound a therapy have not been very adequate [6]. 
A minimal essential for a psychopathology is conflict, a concept 
which is hardly meaningful unless consciousness is posited. But 
conflict is not a sufficient cause of neurosis. Indeed, if our view be 
correct, consciousness is a continual interplay of contending forces; 

* Another patient, a young physicist, reports excessive reading of science 
fiction. "In it," he observes, "the experiments always come out right!" 

Mowrer 89 

and decision making, compromise, and integration are its major ac- 
complishments . 

Where, then, does pathology arise? Section II of this paper gives 
us a clue. There we have seen that there are two broadly different 
principles of feedback. If feedback functions so as to make an 
organism move faster when it lags or slow down when it is going 
too fast, it is said, by the cyberneticists, to be a negative feedback; 
and we are at once reminded of the role of conscience or superego. 
This agency within the total personality is said to be wise, prudent, 
balanced, and mainly concerned with (social) norms. The mentally 
aberrant are said, per contra, to be unstable, unbalanced, abnormal. 
How do they get that way? 

Through parental training as well as by direct experience, human 
beings acquire conscience, which, as we have seen, operates on the 
negative-feedback principle. Other things equal, this should be com- 
patible with the over-all tendency toward self-regulation and home- 
ostasis. However, we have already noted a complication. Man has 
appetites and lusts as well as conscience; and if one operates on a 
conservative, judicious principle, the other is prodigal and reckless. 
Sex is powerfully appetitive, clearly following a positive-feedback 
principle of the "explosive" variety, and anger is likewise the occasion 
for our sometimes "blowing up/' Little wonder, then, that these 
human proclivities are the most difficult to "control" and the ones 
which most often come into conflict with superego functions. Caught 
between two such powerful forces as id and superego, appetite and 
conscience, lust and guilt, what indeed is the "answer"? Here the 
integrative resources of the ego are likely to be taxed to their utmost! 

The Freudian version of the etiology of neurosis holds that in the 
struggle between negative feedback (conscience) and positive feed- 
back (appetite), the former sometimes seizes too firm a control over 
the latter and thus holds down, or inhibits, "instinctual" functioning 
to a pathogenic degree. Anxiety and depression are said to be the 
fruits of this stifling dominion of the superego over id functions [2]. 

On other occasions [4, 5, 6], I have argued for the contrary view, 
that neurosis arises when positive-feedback functions have out-con- 
tended the negative feedback, resulting in temporarily uncontrolled, 
"explosive" behavior for which the individual later feels remorse and 
shame (intensified negative feedback). The ego may then deal with 
this expression of an "aggrieved" conscience in several different ways, 
notably by making a confession and amendments, in which case inner 
stability and harmony tend to be restored; or the ego may set out to 

90 Kentucky Symposium 

"disconnect" the conscience, like the governor on a steam engine, so 
as to let the "wild" behavior occur unobstructedly (albeit usually sur- 
reptitiously). Conscience may be thus dissociated or repressed, and 
the individual may compliment himself on his new freedom, liberty, 
emancipation. But the forces of conscience are tenacious and usu- 
ally come back to haunt their owner, but now not as intelligible guilt 
but as the unintelligible and torturing experiences of anxiety, panic, 
depression, and inferiority feeling. 

It is not my wish or purpose here to debate the relative merits of 
these two interpretations. I present them rather to show how neatly 
and naturally they can both be accommodated in the general theoreti- 
cal framework here presented, one in which the view of "habit" as 
S-R bonds is rejected in favor of a more complicated system in which 
learning is limited to the acquisition of inner meanings which are 
then evaluated and used to arrive at the particular (often novel) 
decision and action which the total situation seems most to warrant. 
One can hardly escape the feeling that cybernetics contains some 
powerfully unifying principles and that science may yet lead to a type 
of synthesis of human experience which has, by some, long been held 


1. Cohn, S. H., and Colin, Sylva M., The role of cybernetics in physiology, 

Scientific Monthly, 1953, 76, 85-89, 

2. Freud, S., A general introduction to psychoanalysis, New York, Liveright 

Publishing Corp., 1935. 

3. Miller, N. E,, Learnable drives and rewards, in Stevens, S. S. (Ed.), Hand- 

book of experimental psychology, New York, John Wiley, 1951. 

4. Mowrer, O. H., Learning theory and personality dynamics, New York, Ronald 

Press, 1950. 

5. Mowrer, O. II., Psychotherapy theory and research, New York, Ronald Press, 


6. Mowrer, O. IL, Neurosis: a disorder of conditioning or problem solving? in 

Miner and Kemp (Eds.), Comparative conditioned neuroses, New York 
Academy of Sciences, 1953, pp. 273-288. 

7. Mowrer, O. H., Is "habit" merely "secondary reinforcement"? (unpublished). 

8. Osgood, C. E., The nature and measurement of meaning, Psycliol. Bull., 1952, 

49, 197-237. 

9. Shannon, C. E., and Weaver, W., The mathematical theory of communica- 

tion, Urbana, University of Illinois Press, 1949. 
10. Wiener, N., Cybernetics, New York, John Wiley, 1948. 

Personality Structures as 
Learning and Motivation Patterns - 
A Theme for the 
Integration of Methodologies 


We have come together at this symposium hoping for an easy and 
happy marriage of two young branches of psychological science; but 
the tragedy of Romeo and Juliet should remind us that the failure to 
consider rivalries of parentage can lead to difficulties even when true 
love exists between the partiesand I am not at all sure that it does 
in this case! Consequently if our purpose is to arrive at joint con- 
cepts and methods we must first face some embarrassing inquiries 
about the implicit viewpoints and ancestries of the parties concerned. 
As a personality theorist, my role at this marriage is perhaps that of 
mother-in-law to learning theory, so your expectations of sweet rea- 
sonableness on my part must be according to your personal projections 
in this situation. 

Quite apart from these immediate differences we have to recognize 
that our more remote common ancestor, psychology, was itself a prob- 
lem child among the sciences, always boasting to answer the spec- 
tacular questions and by polysyllabic global theories before it 
acquired the patience and method effectively to answer the small 
questions about behavior. Theoretically, for example, it indulged in 
such pretentious mathematical-sounding names as topology and in 
such elaborately complete motivational and structural systems as 
those of psychoanalysis,, without first establishing accurate methods 
of description and measurement upon which reliable and worth- 
while laws could be based, or determining personality structure by 
objective, multivariate, factor-analytic methods. Second, it failed to 
develop objective test measures for whatever patterns had been shown 
to be functionally independent or dirnensionally clear. When this is 


92 Kentucky Symposium 

done, and only when, is it economical to begin investigating by learn- 
ing theory the etiology of personality structure. Much valuable re- 
search effort would be wasted if exact workers in learning theory were 
persuaded to explain the development of normal and abnormal per- 
sonality formations as they exist at present in the shaky hunches and 
formulations of clinicians. For though the clinic is admittedly one of 
the best places in which to study personality the implicit multivariate 
analysis which goes on in the clinician's head, when he uses the global, 
wholistic arts of the clinical method, has surely to give way to the 
more explicit multivariate approach of advanced statistics. Thus only 
can the results emerge undistorted by personal prejudice and be sub- 
ject to verification by precise hypothesis testing. 

The personality theorist's position is in many ways intermediate be- 
tween that of the clinician and the learning theorist, in respect to in- 
terests, methods, and theoretical tools. Accordingly, it is up to him to 
indicate, since he stands on the ground where the integration is most 
likely to occur, what difficulties he has in accepting within a common 
system some of the proposals from either flank. 

It is comparatively easy to indicate those encumbrances which I 
think we should not take over from present clinical research. The 
main thing we should not take on is the vast proliferation of mere spe- 
culations, now widely accepted as truths, which occurred before the 
prodigal psychologist repented of his prodigality and recognized the 
necessity of a planned sequence of objective research such as I have 
described. This prodigal phase has been costly enough in research 
time even without our also adopting the debt of its loose conceptual 
offspring. It has filled our clinics with trashy, patent-medicine meas- 
ures of personality, it has littered our journals with researches upon 
the slippery foundations of which more exact scientists cannot build, 
and it has filled the minds of our students with glib, ready-made ex- 
planations where there should be honest doubts and clear perceptions 
of issues open to research. Worst of all it has almost totally filled 
clinical positions with a generation which is largely innocent of scien- 
tific method, which has its interests heavily vested in particular crystal 
balls and test gadgets, and which has no notion, for instance, that a 
criterion is necessary to determine if therapy works at all. Cloistered 
researchers may say that they meet none of these in scientific discus- 
sion, but they are at least social realities. 

Essentially, however, we have turned the corner from this phase, 
thanks to such rigorous measures as the A.P.A/s definition of a clini- 
cian as a fully trained psychologist, and I mention this sad background 

Cattell 93 

only because we have to distinguish in contemporary discussion of 
clinical contributions between what is obviously socially accepted and 
respectable and what can be considered scientifically respectable. I 
shall, therefore, take the position that clinical psychology has con- 
tributed and will continue to contribute a lot of valuable hunches 
about personality for the process of scientific testing. On the other 
hand I cannot take the position that some of my clinical friends 
would like, to the effect that the experimentalist is someone who 
merely checks on clinical hypotheses. For increasingly the valuable 
hypotheses in personality research are arising from experimental work 
itself or at any rate from factor-analytic work. 

The difficulties on the other flank, in relation to learning theory, are 
those between people with similar technical standards but different 
histories. Yet there is one simple respect in which it should be said 
initially that learning theory is more seriously disabled than clinical 
psychology, namely, that whereas clinical psychology has had exten- 
sive empirical contact with personality problems, learning theory 
has had virtually none. Since I have a strong predilection for laws 
that develop out of the phenomena to which they are supposed to be 
relevant, I cannot feel confidence, scientifically, in learning laws taken 
over ready-made from another area, unless and until they justify and 
enlarge themselves in relation to personality measures themselves. 

The proposed ideal research strategy of first determining person- 
ality structure and then determining its laws is therefore embar- 
rassed from the beginning by attitudes and conditions in the history 
of psychology and psychologists, and in this case I favor Oscar Wilde's 
definition of history, namely, "an account of things that should never 
have happened." However, if we may aim even now to institute a 
more ideal and effective sequence I suggest that we recognize five 
steps, or phases, and I propose to list these and devote 10 minutes' 
discussion to them. First there is a phase of precise, quantitative in- 
vestigation of existing structure in the organism, in this case of human 
personality structure [2]; second there is the step of producing reli- 
able instruments for unitary measures of these meaningful structures 
and functions [5]; third and here I ask you to bear with my eccen- 
tricitiesI would ask for an initial investigation of the part played by 
constitution and heredity in these measured patterns and structures, 
so that we do not waste our time trying to explain by learning theory 
what is really a maturational phenomenon. The fourth stage would 
be the application of the usual stimulus-response learning experiment 
design involving sequential, causal, longitudinal analysis of what 

94 Kentucky Symposium 

has first been studied in cross-section, in phases 1 and 2. I say the 
usual learning experiments, but actually this etiological investigation 
would also go beyond controlled experiment. I urge, in addition, new 
multivariate designs for studies in situ, by methods I hope to indicate 
later. Fifth and finally we can expect a phase in which laws of learn- 
ing have themselves become enriched and perhaps modified by the 
new problems and new kinds of material encountered in personality 
change, by a feedback of these augmented principles into the labora- 

Perhaps 10 years hence some psychologist will be able to illustrate 
by pointed examples the principles encountered at each of these 
phases, but my own expansions of them are today very patchy. Re- 
searches directed to even the first two phases have been so little and 
so late that at present one has to turn to articles on which the ink 
is scarcely dry in order to illustrate steps penetrating no further than 
the third phase, and beyond that lies only speculation or work lack- 
ing the foundations indicated. Indeed, as far as I know there are 
just four studies in existence that break into the third phase, that of 
separating inaturational- from learning patterns namely, Eysenck's 
[JO] studies on alleged neurotic patterns, Thurstone's work now in 
progress on primary abilities, a study at Michigan, and the Illinois 
study on personality factors in twins [7]. In this Illinois study we are 
measuring about a dozen known personality factors by objective tests 
on about a thousand identical and fraternal twins, siblings reared 
apart, unrelated children reared together, etc., in order to determine 
nature-nurture ratios for major source traits, as was done some years 
ago for the single source trait of general ability. From this we hope 
a number of definitely learned patterns will clearly segregate to be 
submitted to learning experiment. However, even without the pos- 
sibility of exemplification in phases 4 and 5 it will help discussion of 
the problems therein if we pause to glance more closely at the nature 
of some of the patterns found in phases 1 and 2 [2], for in these fac- 
tors reside the concepts we have to work upon with learning theory. 
Operationally the basis of these patterns is an R-technique factor 
analysis [1] of variables carefully chosen to cover the personality 
sphere, with subsequent discovery of a unique factorial simple struc- 
ture, obtained "blindly" in order to be uninfluenced by the experi- 
menter's theoretical prejudices [6], In this glance at the patterns in 
question I shall not take stock of life-record, clinical-rating data or 
questionnaires, which are properly only preludes to factors in objec- 
tive, situational tests. At least I shall not do so except in so far as the 

Cattell 95 

behavior rating factors permit a matching of the objective -test pat- 
terns with clinical and general criteria. 

It is well known that factors have been obtained, and confirmed on 
new populations, which correspond to the cyclothyme-schizothyme 
dimension of the clinicians, to the concept of emotional maturity or 
ego strength, to the structure of the superego, to general anxiety level, 
and to some further six or more dimensions which differ from those 
just named in that they transcend present clinical insights [2]. 
Though it is encouraging for hopes of integration of clinical and 
quantitative methods to find some degree of convergence of concepts 
it is also not surprising to a methodologist that these finer measure- 
ment and analysis methods yield, as indicated, newer patterns un- 
known to the clinician, just as the microscope yielded species once 
unknown to the biologist. Moreover, although the clinical and factor- 
analytic findings undoubtedly tend to converge, the latter point also 
to important modifications required in the clinical formulations. 

For example, the factor analyses clearly show two factors, A and //, 
not one, in the area ol schizothyme behavior, and one of these, H, has 
physiological, autonomic associations, suggesting some constitutional 
element, whereas the other, A, does not, but has patterns of hostility 
and rigidity suggesting the result of general environmental frustra- 
tions. Further, one can find a factor of paranoid trends, L, quite 
distinct from the schizothyme factors as such, which finding clarifies 
certain clinical classification obscurities [2]. Another instance where 
the clinical picture is both clarified and modified by the factor analysis 
and its reference to normal populations is found in the factor which 
loads the variables concerned with conscientiousness, perseverance in 
maintaining standards, altruistic regard for other people, and other 
manifestations identifying it with the superego conception. This fac- 
tor is virtually uncorrelated with factors of general neuroticism (as 
Mowrer's hypothesis [15] would also require), and its pattern is not 
overly loaded in guilt feelings, which suggests that the clinician's 
view of the superego gained from neurosis formation has unknow- 
ingly been strongly colored by the biased population sample taken. 

As I have already indicated, the methods of the clinician and the 
factor analyst, being wholistic and multivariate, are in a larger sense 
the same, but the clinician uses the analytical powers of his own 
memory whereas the factor analyst prefers the less subjective and 
more refined instruments of quantitative records and calculating ma- 
chines. The important thing initially is that their results agree. But 
it remains for further research, particularly the pioneering of such 

96 Kentucky Symposium 

factor analysts as Ferguson, Wittenborn, and Eysenck [JO], concen- 
trating on abnormal populations, to establish to what extent such dif- 
ferences as I have indicated between factor analyses of the normal 
personality structure and clinical views gathered from the abnormal 
arise, respectively, from differences in objectivity of techniques, and 
from the contrasts of normal and abnormal population samples. 

When now I turn, as promised, to factors in objective personality 
tests themselves I am compelled, in contrast to the discussion of 
rating factors, to ask you to take on trust a weight of evidence which, 
like the bulk of an iceberg, is still largely invisible (and unpublished). 
Since 1944 some seven independent studies, related in a progressive 
sequence converging on gradually clarifying patterns, have been car- 
ried out in our laboratory alone, and only two have yet been pub- 
lished. They cover over 200 different types of personality tests and 
a population of nearly 2000 cases. The upshot is that some seven or 
eight factor patterns have now recurred so consistently as to give us 
considerable confidence in their being very stable source traits in 

By way of illustration we may take that factor [5] which happens 
to correspond to the set of measures which Eysenck found most pow- 
erful in distinguishing neurotics from non-neurotics. It loads most 
highly the tests of sway suggestibility stemming from Hull, the classi- 
cal motor-rigidity or perseveration measures of Spearman, the test of 
attitude fluctuation which we proposed in 1943 [2], and also the 
usual measures of high level of aspiration, of preference for form 
over color, and of a high ratio of the psychogalvanic response to signs 
of threat as contrasted with reaction to the nocive physical stimulus 

If you feel baffled of immediate insights into the underlying prin- 
ciple governing the variables picked out by this factor I beg you not 
to feel worse than the woman in the well-known zoo story who saw 
her first giraffe and remarked, "I just don't believe it." One would be 
disappointed in the method if it turned up only patterns that we 
already know. The method of factor analysis seems to some experi- 
menters peculiar in that it does not initially require that the experi- 
ment start with anything but the most general hypotheses. In fact it 
concurs with the position in Spence's paper, that it is better to begin 
looking for lawful relationships than for verifications of unnecessary 
hypotheses. Factor analysis first presents us with some dimly per- 
ceived pattern, in which initially none of the variables is very sub- 
stantially loaded. But at that point we must form a slightly clearer 

Cattell 97 

hypothesis, by inference and abstraction, from the given variables. 
This is next tested by entering a new factorization with variables more 
pointedly chosen to represent that hypothesis, and so the cycles pro- 
ceed [1, 6, 17]. 

Incidentally, if any personality factor turned out to load very 
highly in our first tests a small set of variables which very closely re- 
sembled one another, I should be suspicious that we had picked up 
some very specific behavior pattern some too narrow group factor, 
in factor-analytic terms. A major personality factor should affect a 
very wide range of behavior, and the factor we have mentioned is 
satisfactory in this respect, lor it covers operational responses which 
can be defined by such apparently diverse operational terms as in- 
stability of attitudes, poor will control, lack of realism, and a tendency 
to project emotional values into stimuli in an irrational fashion. These 
subsume into a single principle only if one takes a sufficiently pene- 
trating theory of personality, but their pattern obviously resembles 
that C factor of ego strength versus emotionality which we have seen 
has been obtained also in clinical and behavior ratings. 

Turning to a second example of these independent factors or source 
traits, we may take the loading pattern [5] comprising slow speed of 
perceptual closure, tendency to perceive threatening objects in un- 
structured pictures, large mean magnitude of general psychogalvanic 
response, absence of questionable reading preferences, high index 
of carefulness in following instructions, and much slowing up of 
reaction time when more complex choice reactions are demanded. 
Here we have hypothesized that the common principle is one of in- 
hibitory tendency, which, when stronger, makes for greater caution 
in coming to perceptual closure, and so on through the other variables 
named. This has been tested to some extent by including in the suc- 
ceeding factorial experiment a test of maze learning under reward 
and under punishment, in which the individuals high on this factor 
were found more responsive to punishment. The whole pattern sug- 
gests that this is the equivalent of the surgency-desurgency factor in 
ratings. But questions of identification with behavior rating factors, 
as well as of interpretations through new variables, cannot be con- 
sidered settled until independent workers have systematically checked 
by further factorization. At present all that we can be sure of is the 
outline of certain factors, and we do have fairly substantial evidence 
that they match certain rating factors and have a certain character. 

Before asking what light the discovery of these patterns may throw 
on the nature of the learning processes through which the organism 

98 Kentucky Symposium 

passes, I must beg time to remind you of a new contribution from this 
direction to the determination of relevant structure, namely, the evi- 
dence about motivation structure [2]. The application of factor 
analysis to motivation and interest variables was initially a forlorn 
hope of structuring that which had hitherto defied objective methods 
as evidenced by the writings of Murray, McDougall, Freud, and 
Tolman. Although the need was great we had grave doubts whether 
the method which had produced a few stable patterns in other realms 
of personality would be able to deal with the complex, subtle, and 
changing relations of this motivation realm. But in fact it succeeded 
beyond our most sanguine hopes [3, S, 9]. 

Fifty diverse attitude-interests had been measured on 200 young 
men by objective methods, that is to say, not by opiuionaire methods 
or self-evaluation but by utilizing the magnitude of the PGR response, 
and by the application of perception and learning principles such as 
retroactive inhibition and attention to cues. The intercorrelations 
of strengths of motivation were factor analyzed and yielded patterns 
that were clearly those of primary drives [2, 3]. 

We had hypothesized that we should find both primary-drive pat- 
terns and sentiments, and had consequently put into the fifty interest- 
attitudes a pair of marker variables, i.e., operational representatives, 
for both drives and sentiments. 1 use "sentiments" here in the same 
general sense as in the accompanying paper by Adams, and like him 
I regard it as of central importance for learning. However, I make 
a slight distinction between sentiments and attitudes in a way which 
he does not [2]. I agree that a chain of attitudes can be recognized 
in which each is instrumental (or "subsidiated" ) to the next, but on 
closer observation it will be found that these interlock in a "dynamic 
lattice" [2]. What I call a sentiment about an object is a point in 
this lattice where several attitudes become integrated [2]. In regard 
to such structures the factor-analytic evidence was puzzling, for we 
obtained evidence only for the pattern of the self-sentiment, that is 
to say, of the integration of attitudes around the self -concept. It 
may be that the covariation of interests centering on such social insti- 
tution patterns as religion, the home, and patriotism is relatively 
slight. Until the study has been repeated a third time, with special 
attention to magnifying this part of the picture, it would, however, 
be premature to deny the possibility of existence of such acquired 
patterns of attitude integration. 

In order to escape that unearned increment of unjustified interpre- 
tations which every psychologist feels entitled to apply to the term 

Cattell 99 

"drives" we have used the term "ergs" for the empirically established 
patterns of hunger, sex, fear or anxiety, dominance, etc., emerging 
from the factor-analytic operations. This term avoids also the current 
semantic confusion of "drives" with "drive." For the present an erg 
means nothing more and nothing less than this immediate motiva- 
tional pattern, and it remains for further research to determine, for 
example, whether the strengths of these nine or more ergs, when ex- 
amined by twin studies or by physiological research, are largely 
innately determined or not. However, their similarity to patterns in 
primates, and in mammals generally, certainly suggests that we are 
dealing with inherited tendencies. Our finding of a very definite erg 
of curiosity or exploration certainly agrees, for example, with the con- 
vincing motivational, problem-solving studies of Harlow [II], who 
shows that monkeys, like professors, will put in a sixteen-hour day 
without such gross rewards as food and drink. Parenthetically these 
researches also support Harlow's questioning of two other pieces of 
semantic nonsense, namely, calling some drives, such as curiosity or 
fear, "secondary" to such "primary" drives as food-seeking or sex, 
and calling all drives "anxiety." I presume that when "instinct" or 
"drive" became socially unpopular someone looked in a dictionary 
and found "tension" was an acceptable synonym for "drive," and a 
little later some second uusophisticate looked in a dictionary and 
found "anxiety" as a bright new synonym for "tension." Anxiety, if 
words are to have any meaning, is surely properly a special derivative 
[2] of the fear or escape drive. But our results clearly show eight 
other drives beside escape, and so, perhaps, following Harlow's apt 
witticism, we shall have to speak of no less than eight "varieties of 
unanxious anxiety." 

If these ergic structurings among objectively measured attitudes 
are confirmed by further, independent researches it will be seen that 
we have made two advances which should be of especial interest for 
learning experiments. First, we have arrived at a pattern of weighted 
elements by which the strength of a given drive or erg can be esti- 
mated in a human being or a lower animal at a given moment in a 
given stimulus situation, and second, we have made possible the 
alternatively oriented analysis of stating quantitatively what particu- 
lar final goals or incentives are concerned in motivating any given 
attitude or habit that we encounter in action in nature. 

The latter method of analysis has been expressed in what I have 
called the ergic principle of attitude measurement. Initially an at- 
titude may be defined as an interest, within a given stimulus situation, 

100 Kentucky Symposium 

in following a certain defined course of action. One may put any 
attitude into the paradigm, "In these circumstances I want so much 
to do this with that/' which defines, as I have indicated elsewhere, 
four conditions: namely, the stimulus; the direction of action taken 
by the response; the object, if any, manipulated by the response; and 
the strength or urgency of the response itself. This basic definition of 
attitude, incidentally, is one which social psychologists have difficulty 
in grasping, a fact which seems to me to have some correlation with 
their polling prediction difficulties, but learning theorists, 1 believe, 
will find it rooted firmly in familiar concepts. 

Now the ergic principle of attitude measurement abstracts from the 
measurement data according to a certain mathematical model pri- 
marily designed to permit a number of useful further calculations to 
be carried out. It expresses the attitude as a vector quantity, the 
length of which represents the strength of the interest in the course 
of action and the direction of which, defined angularly in relation to 
the known ergic coordinates, expresses the quality of the interest, or, 
in other words, the ergic goals satisfied by that course of action. 
Where normative units of measurement can be employed, as in R-tech- 
nique studies, this formulation permits us to add the attitudes of dif- 
ferent people, as, for example, in determining the total attitude of 
members to a group, i.e., its synergy [2] in group dynamics. Where 
ipsative (and normative) units are employed it permits us to work 
out measures of conflict within the individual, of the strength of ergs 
excited in a given stimulus situation, or of the relative investment of 
interest in different learning tasks. 

I spoke a moment ago of the uncertainty we had felt regarding the 
capacity of factor-analytic methods to structure such complex and 
fragile data as motivation measurements, in which methods of ob- 
jective measurement are still tentative and where transient and oscil- 
latory conditions of physiological state and stimulus situation intrude 
into the variance as much as do the fixed characteristics of the situa- 
tion and organism, reducing reliability coefficients on any variable 
measured. But here the method of P-technique factor analysis, which 
we had developed with clinical purposes in mind, actually made capi- 
tal out of the very disadvantages of the situation. When we turned 
it upon the problem, it brought out in still clearer terms the ergic 
motivation structure just described as resulting from the R-techriique 

P-technique measures the same population of attitude-interests on 
one person, day after day, for, say, a hundred days. The changes in 

Cattell 101 

stimulation can either be experimentally arranged or left to the im- 
pact of daily events. Although correlation of the variable series and 
subsequent factorization yielded, as stated [9], the same ergs as had 
been obtained by R-technique studies and also the same self-senti- 
ment structure [8], the exact relationships were now naturally those 
of the unique trait rather than common trait patterns. The personal 
conflicts in the individual, as revealed by the positive and negative 
loading antitheses, when compared with those inferred clinically, 
showed excellent agreement, both as to magnitude and as to the ergs 
and personality structures involved. Further, a calculation, from the 
ergic loading patterns, of the strengths with which ergs were stimu- 
lated, when plotted over the hundred days, showed decided changes 
in level very pertinently related to the events in the young man's 
life. For example, the measured fear or anxiety drive climbed rapidly 
around examination time, and the erg of tender protectiveness (per- 
haps what a rat psychologist would call "maternal drive") rose 
strongly during the week that the subject's father was seriously ill 
in the hospital. Incidentally anyone particularly interested in method 
will have recognized that P-technique, especially if used in what I 
have called the condition-response design [6], may deal with causal 
sequences of stimulus-response and thus transcend the preoccupation 
of ft-technique factorization with cross-sectional relationships only. 
This will be important in our later discussion of methods capable of 
combining the perception of patterns provided by classical factor 
analysis with the perception of stimulus-response relations provided 
by classical itnivariate experiment, as in learning theory studios. 

Now I have taken some of the time up to this midpoint in my talk 
in a detour aimed briefly to describe the foundation of systematic 
conceptions and the methodologies of measurement in personality and 
clinical psychology which in my opinion can be most profitably re- 
lated to learning theory. Such a review has been indispensable be- 
cause much of the work is too recent to be widely circulated among 
specialists in other fields, and yet, without awareness of its outlines, 
it would be impossible to proceed to attempts at exact integration 
with learning theory. As I have indicated above, one cannot be very 
enthusiastic about that variety of learning theory which develops in 
a vacuum as far as personality is concerned and remains vacuous of 
anything but logical constructs invented in an armchair. Would it 
not be better frankly to recognize this as philosophy? 

Now, as suggested a moment ago, the obvious and immediate help 
which personality study can give to learning experiments lies, first, 

102 Kentucky Symposium 

in pointing out the structures which learning needs to explain, and 
second, in providing sound measures by which the increments in 
these structures can be determined. This applies both to learning 
and to that relearning which the clinician calls psychotherapy. In 
regard to this latter it should be noted that the means are now avail- 
able, through basic research in personality, of objectively pleasuring 
personality during changes in psychotherapy. Preliminary results in- 
dicate, as one might suspect on theoretical grounds, that change oc- 
curs not just along one supposed highway of "getting cured" but 
simultaneously with respect to several dimensions of personality. For 
example, in therapy, change occurs in the direction of increased sur- 
gency (factor F), increased integration (factor G), and decreased 
free anxiety (factor O). It is interesting that the changes which 
Petrie [16] has measured in lobotomy follow this same course, with, 
in addition, a decline in factor B or general mental capacity. Some- 
how the therapeutic relearning process produces increased integra- 
tion and decrease of general inhibition other than by the severing of 
neural connections involved in lobotomy. 

This example, dealing as it does with very massive reaction pat- 
terns in the personality namely, those concerned with the degree of 
integration, the general level of inhibition, and the general proneness 
to anxiety responses brings out very clearly the difficulties we now 
have to face in utilizing in personality experiment that learning theory 
which has as yet been untutored by any substantial contact with per- 
sonality theory. Both the personality theorist and the clinician are 
bound to object, if they are frank, that the preoccupation of the learn- 
ing experimenter with "rats and reflexes" has meant altogether insuffi- 
cient regard in learning theory either for the personality patterns 
which are of prime concern or for the psychometric problems of ac- 
curately measuring personality change along such dimensions. For, 
in the first place, the personality factor patterns which make civilized 
man what he is, resulting from his exposure to institutional patterns, 
have rarely or never been investigated or demonstrated in laboratory 
animals. Second, as regards any attempt to measure such patterns, 
learning experiments seem to have been designed on the assumption 
that no attention need be paid to changes in the animal other than 
those in his time of running the maze or acquiring a reflex or whatever 
other single variable has been measured. 

If we are really to get personality and learning theory together, ac- 
cording to the excellent intention of this meeting, I think the above- 

Cattell 103 

implied divergences in viewpoint and method upon which we have 
so far been stumbling only as uncertain innuendoes and embarrassing 
discrepancies should now be explicitly repeated and summarized, and 
this I shall do also under five main headings. As a personality the- 
orist, who has done whatever he is capable of perceiving to rectify 
gaps in personality theory, I shall definitely adopt the position, for 
the sake of clear argument, that the remaining perceptible shortcom- 
ings reside in learning theory. In so doing 1 am not quarreling with 
the position taken in Spence's paper, namely, that learning theory in- 
evitably began with the simplest manifestations and that for the sake 
of morale it was good to get some lawful relations as soon as possible 
at this level. I am only stating as a present fact that learning theory 
has not yet grown beyond that point, and cannot therefore integrate 
as well as one would like with personality theory. But I shall not be 
surprised if the learning theorist, properly aware that his work is in- 
itially well founded, replies, like the Kentucky horse dealer accused 
of selling a beast with defectively short legs, "I see nothing wrong 
with them. They reach down to the ground." But I am merely ask- 
ing whether they can lift the body of personality theory to its proper 

1. First, my criticism that learning theory has concerned itself with 
individual response variables, instead of factor patterns and their 
measurement, may be resolved in part into a charge of neglect of 
the organism as such. Early learning theory simply considered re- 
sponse as a function of stimulus, perhaps with perceptual constructs 
in between, thus: 

=/(S) (1) 

In response to the raised eyebrows of personality theorists some, 
but not all, learning experimenters have formulated response as a 
function of stimulus and organism, thus: 

R = f(0, S) (2) 

but the rigidity of thought is such that this has never been done 
wholeheartedly; and the organism, for many learning theorists, is still 
largely a vacuum, or a disembodied "state," or a rather vague concep- 
tual area populated by intermediate constructs having to do with 
antecedent external stimuli. These concepts are incapable of deal- 
ing with the full facts of individual differences, and they also miss 
some indispensable general characteristics that the general organism 
should bring into the equation. 

104 Kentucky Symposium 

Regarding the first, the learning theorist often takes up a very clear 
position, but it is the position of Wundt in 1880 when he told McKeen 
Cattell that he was interested in common processes, not in individual 
differences. Surely we have learnt by now, however, that the best 
way to find out the nature of common processes is to watch them 
varying, just as the way to perceive a rabbit in a field is to wait till 
it moves. Surely it is understood among us that only such amateurs 
as novelists, applied psychologists, and parents are interested in in- 
dividual differences for their own sake, and that our scientific aim is 
to use them only as the best avenue to general processes and truths. 
Consequently 1 repeat that one of the methodological advances in 
learning experiments that could result from the present contact with 
personality theory would be the effective utilization of individual dif- 
ferences and multivariate methods to identify and measure the in- 
termediate variables between stimulus and response in the above 
equation. In the simpler sense this step would merely introduce into 
the equation the organismic variables, in which the learning theorists 
are to some extent entitled to say they are not particularly interested 
variables such as constitutional levels of intelligence, rigidity, and 
drive strength. (Though I observe that such learning researchers 
as Maier [14] do not share in the general neglect of these factors.) 
But even granting their right only to be interested in the non-organ- 
ismic, non-particular factors, the learning theorists surely need to 
know how the organismic factors interact with the non-organismic 
and what the relative contribution of each is to the total variance in 
the learning process. 

However, what I want to stress is that there is a second applica- 
tion of this individual-difference approach which could yield the 
hypothetical constructs in which learning theory is primarily inter- 
ested, namely, the use of individual differences in previous exposure 
to stimuli, in time intervals between learning experiences, in delay of 
reward, etc. It may be replied with astonishment that these are the 
very substance of the learning theorist's daily bread. But this reply 
misses the point, for two reasons: first, he at present introduces these 
condition variables one or two at a time instead of in a multivariate 
factor-analytic design, such as could yield good definition of factors; 
and second, he proceeds, in analysis of variance and similar designs, 
as if he were quite unconscious that such variables may be correlated. 
I should like to see a closely reasoned discussion of a multivariate de- 
sign to investigate factor analytically the reality of the hypothetical 
constructs in the Hullian equations, such as habit strength, inhibitory 

Cattell 105 

potential, and excitatory potential [12], for the present univariate de- 
signs seem to me to be working with too many unknowns in the ex- 
periment to permit the net of deduction to close in effectively on so 
many constructs from so few variables. 

You will notice, incidentally, that I am indicating theories in terms 
of experimental designs rather than in terms of the familiar theory 
labels, but I do so deliberately because I think we perhaps talk a little 
too much in terms of half-understood generalities and suffer from the 
reification of names that may have no operational justification. Here 
I strongly endorse Adams's remark that we seem to have passed in 
one generation from the barbarism of brute empiricism to the deca- 
dence of an excessive concern with self-conscious theorizing. 

Now 1 should like finally to put before you the personality theorist's 
development of the R = /(O, S) equation and suggest to you that 
the most fertile growing point for theory today would be a side-by- 
side discussion of this formulation on the one hand and the Hullian 
or the Spencean formulation [IS] on the other. If you will permit 
me a metaphor of spring, I am hoping that the seed of the specifica- 
tion equation will become pollinated by a few constants wafted from 
the learning equations. The personality specification equation [2] 
states that a learning performance, R, in a stimulus situation, /, can be 
written thus: 

Rj = SijEi + #2; !r 2 + KsjEa + 84^4 + S MJ 7 T W (3) 

where the S's are situational indices, obtained as factor loadings, and 
the E's and T's are, respectively, the strengths of motivational and 
non-motivational traits peculiar to the organism and its previous learn- 
ing experience. It will be observed that the S's are the dimensions of 
the stimulus situation, giving substance to S in the basic equation (2) 
above, while the E's and T's give the dimensions of the organism, in- 
cluding its past experience, and thus permit a realistic representation 
of O in the basic equation. 

Time shortage, if not more serious shortages, stops me on the 
threshold of discussion of the integrational possibilities. I might point 
out, however, the resemblance in meaning between any one of the 
product terms and the excitatory potential of Hull's scheme, con- 
sidered as a similar product of drive-stimulation strength and habit 
strength. T may be considered the strength of the habit and S the 
capacity of the situation to stimulate drive. Since the personality 
theorist cannot assume that any animal, least of all a human being, 
is acting from one pure motive, the product terms differ from the 

106 Kentucky Symposium 

Hullian formulation here in that they repeat themselves as shown, 
for each of the various established drives in the organism. 

Although I have charged that the first difficulty in getting together 
with learning theorists is that they have left out the organism, which 
to my mind is as bad as producing the play without Hamlet, perhaps 
it is not too late to adopt a Dale Carnegie gesture and put. this crit- 
icism in terms of a much more minor peccadillo. Indeed, if in this 
matter I confine myself to what learning theorists actually do, or 
rather fail to do, instead of to what they think (or what I think they 
think), then the objection would reduce simply to what I have stated 
beforenamely, that they represent the organism, and everything else 
under heaven, by a single response instead of by those patterns which 
the general psychologist, the clinician, and the sociologist regard as 
important. What is worse, they try to infer its nature from this 
single response. I suggest that, quite apart from the interest and im- 
portance of the total patterns, we cannot assume that the laws of 
learning themselves, as they apply to such patterns representative of 
the total organism, are identical with those found for single variables. 

Perhaps it should be made quite clear at this point that because I 
emphasize patterns among responses there is no reason to claim, as 
some do, that I must therefore neglect the relation of responses to 
stimuli. In a well-known paper some years ago Spence criticized [18] 
the methodological defects of those theories and mental measurement 
approaches which seem interested only in relating response to re- 
sponse, whereas the relating of response to stimulus alone, he pointed 
out, gives us those causal sequences in which science is ultimately 
interested. His observation is a penetrating one, and I certainly 
would not stop to defend those item-analyzing educational psycholo- 
gists and psychometrists whose interest finishes at correlation and 
who show no further interest either in learning theory or in general 
personality theory. But I would at least say that as far as I am con- 
cerned the cross-sectional patterns are only preliminaries though very 
essential preliminariesto sequential study.* For good cross-sections 
are as essential to intelligently planned human learning study as are 
well-focused individual pictures to clear perception of action in a 
movie film. However, we do not always have to separate cross-sec- 
tional from longitudinal interests, for there now exist three methodo- 
logical contributions from personality research which inaugurate 

* In these, essentially, we are holding the stimulus constant and seeing how 
varying reactivities produce the pattern to conceive this in classical experimental 

Cattell 107 

simultaneous study of cross-sectional patterns and causal sequences 
namely, (1) P-technique [4], (2) the factorization of increments [6], 
and (3) that combination of factor analysis with factorial methods 
which I have called condition-response factorization [6]. 

2. My second main suggestion for integration concerns the concept 
of ergs and the methodology of measuring drive strengths. For I 
trust that the introductory illustrations suffice to support the argu- 
ment that we need to shift from single variables to patterns not only 
in terms of response measurement but also in terms of measuring the 
manifestations of the independent variable the drive strength. It is 
inadequate and unnecessary to restrict the measurement of thirst to 
hours of water deprivation or of fear to the measured voltage of a 
shock. The work I cited shows that with human beings the number 
and the nature of independent ergs can be determined, and that one 
can objectively discover the weights by which the strength of excita- 
tion of a drive can be estimated from a measured set of manifestations. 
Anderson's pioneer work on factoring the reactivities of the rat points 
to the practicality of similar substitutions of pattern for single vari- 
ables there also. 

If we are to be scientifically cautious we must admit that the learn- 
ing curves and laws obtained for differences of ergic strength, as thus 
more completely conceived and measured, could be significantly dif- 
ferent from those so far obtained on a univariate basis. Even if they 
do not turn out to be different in general form, as some of Harlow's 
observations on animals suggest, they will almost certainly be more 
accurate. Furthermore, such an approach could open up the whole 
question of whether the differences of quality between drives are as- 
sociated with important differences in the process of learning and 
forgetting. For example, it is implicit in Freud that neurosis and re- 
pression occur only with appetitive, viscerogenic drives; and Brozek's 
demonstration of human experimental neurosis through hunger still 
leaves positive evidence only for the efficacy of this same drive cate- 
gory. Is it possible that forgetting is a different kind of process with 
non-viscerogenic drives such as fear, curiosity, and gregariousness? 
With objective delineation of the drive structures and the determina- 
tion of weighted composites by which they may be measured, re- 
search on this question becomes possible. 

3. Next, in the discussion of rapprochement, I want simply to ask a 
question rather than make a suggestion, for here the learning field 
seems too semantically confused to make any intelligible suggestion 
without much time spent on defining terms. The question is this, 

108 Kentucky Symposium 

"How can we get together when personality theorists, following clini- 
cians, conceive learning as a dynamic process, in terms of goals and 
rewards, whereas for a substantial - number of learning theorists the 
paradigm of learning always seems to be primarily the conditioned re- 
flex?" I suppose the answer is that we can get together only with 
learning theorists like Hull, Thorndike, and Spence, who bring con- 
ditioning under reward learning, or with two-factorists like Skinner, 
Mowrer [15], and Maier. But if the position of the latter should 
prove to be scientifically sound then the personality theorists and cli- 
nicians are grievously neglecting a principle that must have more 
representation in their data than they have yet recognizedand per- 
sonality theory will have to be reformed accordingly. 

4. My fourth suggestion regarding our integration difficulties is 
that we should be prepared to admit in our diagnosis that the gaps are 
sometimes wide because they are historically old and deep. Let us 
be clinicians and try to remove the dissociations by facing our past 
traumata. Historically, in the nineteenth century, the pukka sahibs 
of scientific psychology were unquestionably to be found in the labor- 
atory, rather than in the clinic, and within this shrine their fetish was 
the brass instrument and the univariate, controlled experiment- 
sacred relics from physics. To this day many psychologists in percep- 
tion and learning seem not to have realized that this particular ritual 
is not the whole of scientific method but only one narrow branch of it. 

Now I have no intrinsic objection to this scientific "holier-than-thou" 
attitude so prevalent among learning theorists if it would not inter- 
fere with their learning. Parenthetically I have no doubt that they 
are indeed more scientifically respectable than, say, psychoanalysts, 
and their attitude is better than the "more-popular-than-thou" atti- 
tude of the clinicians or the "richer -than-thou" attitude of industrial 
psychologists. But what these scientific pharisees or hermits (accord- 
ing to your view) fail to realize is that in many investigations the new 
multivariate methods, which psychology itself may justifiably feel 
proud to have given to the sciences, have better claims to scientific 
penetration and rigor than the imitations of physics with which the 
old guard has so rigidly been practicing. 

To all but these Rip Van Winkles it has been evident for a genera- 
tion that important human learning situations, such as those with 
which clinical and social psychology have to deal, cannot be dragged 
intact and alive into the laboratory. The controlled, univariate experi- 
ment, in which nothing but the independent variable alters to an 
important degree, becomes inapplicable and obsolete if only because 

Cattell 109 

of the limits to our rights of major interference in human lives. If at 
that point you can think of nothing better and you prefer the shadow 
of scientific method to the substance of living psychological events, 
you can continue the univariate experimental procedure with animals; 
but this, as I think Mowrer has found, is a cul-de-sac which is soon 
exhausted, for you cannot reason very reliably by analogy from rats 
about human social problems. An alternative scientific method had 
therefore to be conceived, in which we let events happen in life as 
they will and tease out by statistical finesse what cannot be handled 
by brute experimental control. We are indebted to such men as 
Spearman [17], Fisher, Thurstone [19], and Burt [1] for most clearly 
perceiving this need and starting to hack a new path away from the 
beaten but misleading track of the classical experiment. 

But multivariate statistical designs do more than provide an effec- 
tive way of handling what used to be called the controlled variable 
and the uncontrolled variable in situations where control is impossible. 
The factor-analytic approach [6*], for example, in addition tells us first 
where much of the unknown error variance is located, and second, 
much more about the operations by which our concepts are defined. 

We must look at this last point more sharply, for it contains the 
crux of an issue which is deeper than that of experimental design alone 
and which has been hovering in the background of my earlier points. 
May I assume that we are all good operationists together and that 
the established meaning of a concept goes only as far as the operations 
invoked to represent or to test it? Now the univariate experiment 
leaves the meaning of a concept hitched to a single operation. For 
example, hunger is hours of deprivation of food and emotionality in 
a rat is the number of defecations in an open space. I submit that 
when the sharp winds of argument about "meaning" begin to blow 
it is much better to have a concept tied down at the corners by sev- 
eral variables than by some isolated operation. When a personality 
dimension or motivation strength is hypothesized as a whole pattern 
of operations of measurement, not only is it more effectively defined 
and more conclusively tested by experiment but it also provides a 
firmer basis for an architectonic building of further research, as a 
table with several legs is a firmer foundation than a one-legged table. 

Or, to look at this point another way, there are an infinite number 
of individual variables by which one might choose to represent some 
concept such as rigidity or ego strength or anxiety or drive strength. 
History unfortunately shows that no two successive studies will use 
the same variable for the same concept, for the modern psychologist 

110 Kentucky Symposium 

might lose his individuality if not his soul by such obvious lack of 
originality. Indeed the capacity of a finite number of psychologists 
to spread themselves over an infinite number of variables is something 
for sampling theorists and even mathematicians to marvel over. But 
the well-known upshot of this marvel is that no two studies integrate, 
and that where chemists make architectonic progress, by deigning to 
recognize the same elements, psychological research is all too fre- 
quently circular or inconclusive. If psychologists would deal with 
established patterns among variables in a given area, as their ref- 
erence concepts in research, this difficulty would vanish. 

An illustration of what happens in this respect when univariate 
learning theorists get loose in the multivariate realm of personality is 
provided by half a dozen attempts to do laboratory experiments on 
the relation of such matters as the self-concept, or personality levels 
in anxiety, to such laboratory measurements as rate of extinction of 
reflexes, rate of closure, or speed of reaction time. The contrast be- 
tween the impeccable scientific design developed in the laboratory 
measures and the uninformed plan for the measurement of anxiety is 
a painful one. If we are to build bridges from personality to learning 
theory why not build them expertly at both ends? In these cases the 
learning theorists have founded the bridge on an admirable pier of 
solid experiment at their end, but have constructed the other pier 
that concerned with defining and measuring anxiety out of an ama- 
teurish heap of rubble. They have had no better conception than to 
throw together sets of questionnaire items having "face- validity," 
which, as we all know, is a polite term for no validity at all. I submit 
that an examination of the evidence on personality factors would have 
indicated the following three points: (1) that though available data 
show only one erg of fear or escape [3, 9] there are indications that 
its manifestations become separated into four distinct sources [2] in 
personality structure, namely, those covered by the factors labeled 
F, M, O, and Q4. As far as interpretation is now possible these cor- 
respond to an anxiety level of general previous punishment, an anxiety 
of superego action, a free-floating "anxiety hysteria" anxiety, and a 
somatic anxiety like that defined in anxiety neurosis. Almost certainly 
the associations of the composite with conditioning are due to only 
one of these, and the researches in question will need to be done 
over again to show which one, and to make an intelligible interpreta- 
tion possible. Incidentally, the extraordinary semantic carelessness 
which says anxiety level is only drive level would still further confuse 

Cattell 111 

the design and interpretation of such experiments: the fact is that 
the questionnaires used have dealt with anxiety, not drive; (2) that 
furthermore the above four independent forms of anxiety in person- 
ality are in some cases measurable by more objective devices than the 
questionnaire; and (3) that some indication could have been ob- 
tained from previous work, notably the observations of Stephenson 
on the resemblance of extinction phenomena to the classical motor- 
rigidity factor measures, and the extensively investigated relations of 
this rigidity factor to personality, as to which personality factors, 
anxious or non-anxious, are more likely to show significant relations 
to conditioning extinction. 

5. I think that two of the four just causes and impediments to the 
wedding of learning and personality theory that I have pointed out 
may be considered to be theoretically complex and debatable, and 
the other two obvious and remediable. My fifth and last impediment 
is, I hope, simply remediable though theoretically complex. This par- 
ticular obstacle to bringing personality theory and learning theory 
together resides in the fact that in human learning the rewards are 
often so obscure, so rooted in unconscious rewards (such as those to 
superego value systems), or so hidden in faint cues in the social en- 
vironment, that any attempt to relate drive strength to learning in 
such situations, in the honest fashion of rat and laboratory experi- 
ments, appears to be denied a chance. To add to the complication 
the proportions of the standard 'Varieties of learning" in these human 
fields of experience are in general very different from those to which 
we are accustomed in the laboratory. On the question of varieties 
of learning [2] I presume we can agree descriptively on three: (1) 
conditioning by contiguity, (2) rewarded means-end learning, and (3) 
rewarded "integration" learning though some would run the second 
and third together. However, though they are both forms of dynamic 
learning, the first of them concerns a single drive and only changes in 
the means to its goal attainment are involved, whereas in the second 
there is the larger problem of conflict resolution. This integration 
learning, moreover, is the principal concern of the clinical psycholo- 
gist. Briefly it may be defined by the fact that conflict ends in the 
denial of some individual drive goals in the interests of a greater re- 
ward to the total needs of the organism. 

Now the specification equation analysis of any given learning per- 
formance into situational indices on the one hand and ergic and 
metanergic personality structures on the other as stated earlier 

112 Kentucky Symposium 

actually provides us both with the information we need concerning 
the sign-significate satisfactions existing in the obscure total stimulus 
situation and also with a statement of whether the learning is largely 
a means-end learning or an integration learning. The representation 
of the first requires no amplification: it has already been seen that the 
S's define the motivational dimensions of the stimulus situation. Re- 
garding the nature of the learning I shall simply point out for further 
reflection that the situation involves integration learning to the ex- 
tent that there are conflicting positive and negative loadings to the 
situational indices, particularly in respect to the self-sentiment and 
primary drives. A difference of sign means that a positive satisfaction 
of one erg is to be obtained only at the cost of suppression of another. 
Some interesting examples have been discussed elsewhere [8, 9]. 

Regarding the revelation of the cues that lie in the situation, it has 
just been stated that magnitudes of the situational indices attached 
to particular drives indicate (with certain statistical modifications) 
the extent of the provocations to those drives somehow hidden in the 
situation. The next step in proceeding from this basis would be to 
locate and manipulate the specific stimuli or aspects of the situation 
concerned, whereupon we might hope to obtain, by the ensuing learn- 
ing experiment, an experimental check on the factor-analysis S values. 
A methodological alternative would be to accept the strengths of ex- 
citation given by the factorization and attempt to establish laws of 
learning by relating the emergent personality patterns to these ac- 
cepted ergic strengths. 

In conclusion it is my impression that I have rarely seen different 
specialist fields brought together in which the hopes of substantial 
gains from interactions are so great. There exist at any rate two sub- 
stantial and precise predictive systems: first, the crystallization of facts 
and laws in factor-analytic measures of personality, and second, the 
equally effective crystallizations in learning theory. Since both are 
firmly rooted in experiment there must be some way of linking one 
with the other in a further set of lawful relationships. For researchers 
who aspire to make those links there now exist some potent sugges- 
tions in methods and hypotheses, which can be utilized by those bold 
enough to think in new ways and patient enough to study the the- 
oretical implications of what has been offered. One reason why so 
many possibilities have been unearthed here is that this meeting of 
specialists has been long overdue, and the University of Kentucky 
is to be congratulated on the initiative of its psychology department 
in at length bringing this synthesis about. 

Cattell 113 


1. Burt, C., Factors of the mind, London, University of London Press, 1940. 

2. Cattell, R. B., Personality: a systematic, theoretical, and factual study, New 

York, McGraw-Hill, 1950. 

3. Cattell, R. B., The discovery of ergic structure in man in terms of common 

attitudes, J. abnorm. soc. PsychoL, 1950, 45, 598-618. 

4. Cattell, R. B., P-teehnique, a new method for analyzing the structure of per- 

sonal motivation, Trans. N. Y. Acad. Set., 1951, 14, 29-34. 

5. Cattell, R. B., A factorization of tests of personality source traits, Brit. J. 

PsychoL, Stat. Sect., 1951, 4, 165-178. 

6. Cattell, R. B., Factor analysis for psychologists, New York, Harper and 

Brothers, 1952. 

7. Cattell, R. B., Research designing in psychological genetics with special ref- 

erence to the multiple variance method, J. human Genetics, 1953, 5, 121, 

8. Cattell, R. B., and Miller, A., A confirmation of the ergic and self -sentiment 

patterns among dynamic traits as determined by K-technique, Brit. J. 
PsychoL, 1952, 43, 280-294. 

9. Cattell, R. B., and Cross, K. P., Comparison of the ergic and self-sentiment 

structures found in dynamic traits by H- and F-techniqucs, J. Pers., 1952, 
21, 250-272. 

10. Eysenck, H. J., The scientific study of personality, New York, Harcourt, Brace, 


11. Harlow, H., see his article in this volume. 

12. Hull, C. L., Mind, mechanism, and adaptive behavior, PsychoL Rev., 1937, 

44, 1-32. 

13. Hull, C. L., Principles of behavior, New York, Appleton-Ccntury, 1943. 

14. Maier, N. F., Frustration: the study of behavior without a goal, New York, 

McCraw-IIill, 1949. 

15. Mowrer, O. H., Learning theory and personality dynamics, New York, Ronald 

Press, 1950. 

16. Petrie, A., Personality and the frontal lobes, London, Routledge and Kegan 

Paul, 1952. 

17. Spearman, C., Psychology down the ages, London, Macmillan, 1937. 

18. Spence, K. W., The nature of theory construction in contemporary psychology, 

PsychoL Rev., 1944, 57, 47-68. 
J 9. Thurstone, L. L., Multiple factor analysis, Chicago, University of Chicago Press. 

Prospects and Perspectives in 
Psychotherapeutic Theory 
and Research 


It is quite apparent to anyone who reads the literature on psycho- 
therapy that the level of discourse is low and that the different schools 
of psychotherapy have different systems of propositions. These sys- 
tematic differences make intercommunication with regard to single 
propositions extremely difficult and in so far as science involves good 
intercommunication demonstrates the low level of scientific under- 
standing in this domain. It will be my purpose in this discussion to 
consider how, in a domain in which all seem to profess a scientific 
orientation, this low level of communication came into being and to 
consider how the domain of learning can serve to provide a common 
frame of reference which would promote communication as between 
systems of psychotherapy. 

In reviewing the modern history of psychotherapy, which as far as 
I am concerned dates from the early days of Freud and Janet, he who 
has the advantage of hindsight and contemporary scientific sophistica- 
tion sees a fascinating spectacle. He sees a host of brilliant men in- 
spired by the success of physical and biological science, bending their 
energies to the creative formulations of the processes of psychotherapy 
and of the nature of human nature. These men, who worked in a re- 
stricted and really rather simple social situation, that of psycho- 
therapy, which had the unique feature of being safe for the therapist 
in a way in which everyday life was not, were therefore in a position 
to make observations which were relatively objective, i.e., were not 
ego-involved. And they were, it should be noted, with a given indi- 
vidual for a considerable span of time. Today it is clear enough 
that, although these circumstances permitted relatively objective ob- 
servation, still the behavior of these therapists was in itself influential 
in producing the phenomena (and their time sequences) from which 


Butler 115 

theories of psychotherapy and, inevitably, theories of human nature 
arose. It is also clear enough that no two of these men were identical 
or exhibited the same behavior pattern in therapy. 

Consequently it seems quite evident, to one who is looking back, 
that different observations would be made, and that different formula- 
tions of psychotherapy and personality theories would emerge. It 
is also to be expected that these men would make observations and 
evolve theories which would be quite unacceptable and upsetting 
to their contemporaries. And indeed it is true that even now psycho- 
therapists are often subject to severe attacks because of the threat- 
ening nature of their conclusions. Seen in perspective then, it is quite 
natural that these men who believed in the accuracy of their observa- 
tions, who were convinced of the importance of their work, and who 
were the objects of severe attack, would be intolerant of opposition 
both from the public and from their colleagues who, on the basis of 
different observations, differed theoretically. 

The basic point here is that in their time no basis for settling the 
to-be-expected differences existed, even in principle. These men 
were endeavoring to take a scientific attitude in a domain in which 
very few, if any, scientific studies were possible. What were the 
techniques applicable to the scientific study of human behavior in 
1900, for example? And, for that matter, where existed the concept 
of the study of behavior in 1900? What theory of science existed in 
1900 which would help the therapist to even formulate a scientific in- 
vestigation of his problems? It is certain that Pearson's "grammar of 
science" could not aid the therapist; neither could the then current 
psychological theories (though they could and did contribute to ther- 
apeutic technique) nor the discipline of statistics. 

We expect then, under such circumstances, to find these able and 
gifted men differing deeply in their approaches and theories, con- 
vinced of the general importance of their work with little in the way 
of scientific methods to aid them in settling their differences. And 
we also expect them, in such circumstances, to settle their differences 
in the ancient ways: by dismissal, by undermining, by expulsion from 
the group, by leaving the group, and, above all, by impassioned per- 
suasion. It is clear that for the most part such social processes de- 
termined the acceptance and rejection of therapies at least as much 
as their validity. Who is to say whether "individual psychology" is 
better or worse or more valid than psychoanalysis, on the basis of 
scientific evidence? Yet it is fairly clear today that Freud left more 
intellectual heirs than Adler, and that as a group enterprise psycho- 

116 Kentucky Symposium 

analysis is more successful than "individual psychology" in terms of 
number of publications, publicity, and adherents, both practicing and 
devotional. As for those who did not join groups, did not have the 
temperament for scholastic disputes, who remembers them? 

As we look to the theories evolved we see that some are more con- 
sistent than others, some seem to make more sense than others, and 
some have broader scope than others. But as we look at the evidence 
we see what are called in the area of student personnel work "system- 
atic anecdotal records." We see accounts of the behavior of patients 
fitted into pre-existing frameworks with certain hours picked out for 
analysis, with an implicit rather than explicit coding of behavior; we 
see perhaps hundreds of therapeutic hours condensed into a few 
pages of description, followed quite often by really impressive the- 
orizing which has no ascertainably direct connection with the basic 
data, i.e., with the coded bc\havior of the patient or patients. And 
this condition has persisted, with but very few exceptions, to the 

With this perspective of the past it is hardly surprising to find 
certain kinds of vested interests developing over a rather considerable 
time period with as some of their consequences a general lack of in- 
terest in evidence, strongly grounded beliefs in the truth and validity 
of one kind of theoretical approach, and a general circularity of theory 
which is most discouraging to those who really want to approach the 
domain of psychotherapy with both a scientific attitude and a scien- 
tific method of implementing that attitude. 

To be quite frank, it is my opinion that, dark as the perspective is, 
the present is quite as dark when one views the different extant schools 
of psychotherapy. What once represented an attempt at scientific 
formulations, or at least some kind of prologue to scientific formulations, 
now represent dogmas, now are limiting frames of reference, the 
boundaries of which it is dangerous to approach if one values his 
standing in his group and is unwilling to risk or to undertake the 
formation of a new group. The darkness of this picture of parochial- 
ism and intellectual provincialism is aggravated when one considers 
that today there are impressive general theories of science, highly 
sophisticated methods of statistical analysis which are consistent with 
scientific method, and a growth in psychological theory and experi- 
mentation which is largely ignored by psychotherapists who still 
grind out books on human nature, claim one theory is "better" than 
another, and claim priority for the validity of their theories of per- 
sonalityall in the name of science. For myself, I see no way out of 

Butler 117 

this situation from within. It seems to me that the dilemma- must be 
solved from the outside, that the parochialism can only be broken 
down by reducing theories of therapy and therapy-based personality 
theories to more general psychological theories which will "contain" 
all of them, so to speak. The apparent contradictions between the 
propositions put forth by the different therapeutic schools can only 
thus, it seems to me, be really examined. And it seems to me to be 
almost certain that the therapists are not going to do this themselves. 
At least they show few signs of wanting to. 

I am of the opinion, at the moment at least, that this aim can best 
be realized through learning theory. Admittedly, learning theory is 
parochial in the sense that there are several theories of learning. 
There is a decided difference, however, between parochialism, if this 
is indeed the proper word to use, in the domain of learning and 
parochialism in the domain of psychotherapy. Students of learning 
have committed themselves to a common enterprise, the scientific 
enterprise. Therefore, learning theorists are compelled to listen to 
each other. Tolman cannot dismiss Spence's theories as superficial 
even if he wanted to. He must and does address himself to experi- 
ments supporting or refuting Spence's claims, and Spence must do 
likewise. And over the years both behavior theory and expectancy 
theory have accommodated themselves each to the other on the basis 
of experimental results. What we have seen develop is both a cer- 
tain convergence and an increasingly clear discussion of crucial the- 
oretical differences. 

With therapeutic theory the situation is far different. So far none 
of the major therapeutic theorists appears to have felt compelled to 
notice his major rivals except to dismiss them as superficial and ob- 
viously incorrect when indeed they have not been ignored. Their 
parochialism rests on dismissal, rejection, and the assumption of 
biased and superficial observation. 

I am not, therefore, disturbed when a behavior theorist or an ex- 
pectancy theorist or some other learning theorist decides to translate 
propositions in the domain of psychotherapy into terms of the particu- 
lar learning theory of his choice. The propositions of learning theory 
are, by and large, grounded in a theoretic-experimental system. I 
am not even disturbed if the theorist extends his theory in a consistent 
way, without experimentation, in order to accomplish this end. I am 
disturbed when he decides to restrict his consideration of the psycho- 
therapeutic domain to client-centered therapy or "individual psy- 
chology" or "analytic psychology" or "conditioned reflex therapy" or 

118 Kentucky Symposium 

"dianetics" or what not, because this or that therapy has "unques- 
tionably contributed the most" or "I am most familiar with it" or "It 
is easiest to translate." 

In my opinion such restrictions perpetuate the existing parochial- 
isms in the domain of psychotherapy. Either conversion takes place 
because coordination or translation is confused with reduction, or 
an impressive step forward is rejected because propositions favored 
by one's own variety of therapy are apparently invalidated by the 
translation. Therapists are not prone to give up formulations based 
on personal, intimate, and hard-won experience. 

Although obvious enough, this matter of translation should per- 
haps be discussed explicitly. When a learning theorist is able to find 
a certain consistency between the propositions of a given variety of 
psychotherapy and his own variety of learning theory, he obviously 
regards his learning theory as being better in the sense of science 
than the therapeutic theory he is translating. Otherwise the transla- 
tion would be in the other direction. I assume then that what hap- 
pens is that the relatively inexact propositions of one domain, psycho- 
therapy, are being coordinated with the relatively exact propositions 
of another domain, learning theory. Now I would not deny that 
such translations are potentially worth while; in fact, I assert that 
they are. There are, however, two considerations to be kept in mind: 

1. The translations contribute nothing to our knowledge of psycho- 
therapy. The specifiable referents of the terms of learning theory 
reside in the experimental situations created by the learning experi- 

2. The behavior of patients and clients and therapists in psycho- 
therapy is as unspecified as ever it was, and this is very unspecified 
indeed. What are exact terms in learning theory then become as 
inexact as their semantic equivalents in psychotherapeutic and per- 
sonality theory, and the connections between the propositions are 
connections which can be demonstrated to hold for one domain but 
cannot be demonstrated to hold for the other. To put it bluntly, the 
unification of the domains has by and large been on the semantic 
level. Once one has succeeded in finding a certain isomorphism be- 
tween the propositions of therapeutic theory, the propositions of 
therapy-based personality theories, and the propositions of learning 
theories, with some paring down of the former propositions and some 
relaxation of the latter, one finds himself turning to experiments with 
animals and finding analogues. These analogues, I might add, have 
not contributed much, if anything, to learning theory. Why should 

Butler 119 

they when the experimenter started with learning theory, found 
isomorphisms, and came back to the domain of learning? They do 
not contribute directly to psychotherapeutic theories because the 
referents of the terms used in such theories are unknown in any scien- 
tific sense; if any direct contribution is to be made it is to be made to 
learning theory. 

Perhaps I can illustrate these points by citations from current litera- 
ture. Dollard and Miller [2] have written a most stimulating and 
provocative book entitled Personality and psychotherapy, by which 
is meant apparently behavior theory and psychoanalysis. On p. 257, 
unsurprisingly, we find that "therapy [psychoanalysis] takes time/* 
Why? Because (1) anxiety is produced by free association; (2) fac- 
ing anxiety is painful and exhausting; (3) a self-chosen pace (in psy- 
choanalysis) is a slow pace, and as a result extinction of anxiety 
attached to verbal responses is a slow affair; (4) the therapist can 
best advance therapy by insisting on free association. 

I take it that these statements can be traced back to the propositions 
of their approach to behavior theory if "anxiety" is accepted as a 
strong stimulus. Thus: 

If anxiety is a strong stimulus, 

if anxiety is produced by free association, 

then any response or classes of responses which reduce anxiety 
will be reinforced. Thus free associating will tend to cease. 

If the therapist insists on free association, 


if the free associating arouses anxiety, 

then the therapist becomes a set of cues which arouses anxiety. 

I find this if-then sequence which seems to be implicit in Dollard 
and Miller's account to be entirely reasonable. I simply want to 
emphasize, which they do not, that the therapist in his behavior both 
reduces and arouses anxiety and that thus, by his own behavior, re- 
gardless of transference, arouses directly competing response tend- 
encies. And this may go on simultaneously; the therapist may be 
reassuring, etc., in his manner while insisting on free association. It 
follows from this that extinction of extinction is going on in their 
description of psychoanalysis with the balance being on the side of 
anxiety reduction in successful cases. 

Now take a proposition apparently contradictory to the one that 
"therapy necessarily takes [a long] time"; i.e., "therapy does not 

120 Kentucky Symposium 

necessarily take [a long] time." The latter statement will be recog- 
nized as rising out of client-centered therapy. Why is it that therapy 
does not take time? Because the client-centered therapist attempts 
by his behavior to create a "safe, accepting, non-threatening atmos- 
phere," in which the only rules for client behavior consist of limits 
rather than specific directions. In terms of behavior theory: 4 

If anxiety is a strong stimulus, 


if a given instrumental (verbal) behavior is a cue for anxiety, 


if the understanding and accepting behavior of the therapist 

(a) rewards the instrumental behavior (higher-order reinforce- 

(b) reduces anxiety (second-order reinforcement), 


( 1 ) by gradient of reinforcement the immediately preceding covert 
and overt responses are reinforced, 

(2) by generalization similar covert responses or response tend- 
encies are reinforced which produces a decrement in their 
tendency to evoke anxiety. 

We now assume that the instrumental behavior reinforced is less 
anxiety-producing than the immediately preceding covert response 
( def ensiveness ) . This is consonant with the position of Dollard and 
Miller. Then the immediately preceding covert response is hypothe- 
sized to be (from gradient of reinforcement and drop in cue value 
for anxiety) prepotent over the verbal response just made. If not the 
first response would be repeated. 

It follows from the above analysis that: 

1. Progressive motivation is provided to express "dangerous 
thoughts," since those in the response sequence become progressively 
less "dangerous" than they were but retain their prepotence over other 

2. By generalization "similar" thoughts become "less dangerous," 
and since some of the responses similar to verbal response 1 are simi- 
lar to verbal response 2, which in turn are similar to those of response 
3, multiple sources of reinforcement and loss of cue value for anxiety 
of unexpressed thoughts "similar" to those of the expressed thoughts 

Butler 121 

3. "The self-chosen pace of the client is a slow pace" is correct only 
in so far as the therapist arouses anxiety as well as reduces it, as Dol- 
lard and Miller point out but do not exploit. 

At any rate, this analysis puts the issue just where it belongs. The 
two contradictory propositions, "therapy necessarily takes [a long] 
time" and "therapy does not necessarily take [a long] time/' are both 
consistent with behavior theory. The contradictions in the proposi- 
tions really reside in what is regarded as necessary behavior on the 
part of the therapist. This analysis in Bollard's and Miller's own 
terms shows that "dangerous" thoughts become less dangerous both 
directly and indirectly, and these thoughts should be expressed when 
the role of the therapist as an agent for anxiety is being minimized. 
Also, since the anxiety evoked by the unexpressed thoughts receives 
increments of extinction time after time, it may be seen that topics 
never mentioned or even "thought about" may be dealt with "ade- 
quately" in therapy, i.e., may become "conscious" and be dealt with 
in terms of anticipated consequences rather than reacted to indirectly 
by way of anxiety reduction (avoidance). 

Finally, the problem of "time" in therapy has thus become an em- 
pirical problem, and the solution lies in the study of client or patient 
behavior in the presence of various constellations of therapist be- 
havior. This conclusion is, of course, true only if behavior theory of 
the Dollard and Miller variety is considered to be a valid theory. For 
myself, I can see as much reason for accepting this brand of behavior 
theory as I can for accepting any current theories of psychotherapy 
and of personality. The test really lies in the study of therapeutic 
behavior, client and therapist, not in the learning theory. One unique 
function of learning theory can be to connect the apparently incon- 
sistent or contradictory propositions of therapeutic theory and to 
direct attention to the consequences of therapist behavior, timewise 
and otherwise, in promoting adjustment. In other words learning 
theory will, I think, if properly used, emphasize the conditions of 
learning improved by the behavior of the therapist and suggest ex- 
planations of the resulting behavior of clients in terms of common 

As yet little has been done by learning theorists to relate behavior 
theory to psychotherapeutically based personality theories other than 
to equate the ego with the higher mental processes and the id with 
drives [2, 5], Perhaps this has been because expressions such as 
"self or "ego" seem to denote some kind of entity, some "little man 
up there." My clinical experience, however, leads me to believe 

122 Kentucky Symposium 

that, when we clinicians speak of selves and egos or self-processes or 
ego-processes, we are speaking not of entities and not of processes 
considered as such but of the organization of the higher mental 
processes, of systems of processes, or unities which are open systems. 
The clinicians, in implying that these systems are open, mean that 
these systems are subject to change; that reorganization of the system 
may be going on more or less continuously and that in some sense 
disorganization may be present, especially when the change in the 
system is great. 

Now, if the higher mental processes (thoughts and images) are 
considered responses, as Dollard and Miller so consider them, they 
must like all other responses be (or generate) stimuli. That is, they 
are response cues, and we may consider them as subject to generaliza- 
tion, extinction, etc. that is to say, the propositions of behavior 
theory may be applied to them. 

When we consider such functional stimuli and functional responses 
it is clear that they cannot be defined along physical or similarity di- 
mensions like actual stimulus and response generalization. Further- 
more the distinction between response and stimulus generalization 
which is possible in the non-functional definitions of the dimensions 
no longer seems warranted. Response generalization and stimulus 
generalization must mean the same thing, since for functional re- 
sponses and functional stimuli the response is identically a stimulus. 
Furthermore, it would seem that since thoughts (functional cue-pro- 
ducing responses) are acquired by experience, by reinforcement, the 
generalization dimensions are themselves functional, i.e., are learned. 
For example, the inner response symbolized by "I am short" may 
connote "I am inadequate." The two responses are on the same 
generalization dimension or gradient. I think that this notion is en- 
tirely consistent with the position of Dollard and Miller, if, indeed, 
it is not part of their position. 

If the notion that generalization of symbolic processes is based on 
experience, that the generalization dimensions are functional, is 
acceptable, then a given symbolic cue-producing response may lie 
on more than one generalization dimension. A thought may lead to 
several other thoughts which among themselves may be independent. 
Take, for example, a given thought, A, evoked by external stimuli, 
which is on three generalization dimensions, 1, 2, 3. Then A may 
evoke B, C, and D, which are on dimensions 1, 2, and 3, respectively. 
B, however, will not evoke C and D except by way of A. We may 
represent this situation by considering each generalization to be an 

Butler 123 

arrow where the arrowhead represents the response highest on the 
gradient. Then A is at the point of intersection of the three arrows. 
It should be noted that the point of intersection need not be at the 
center of the arrows. A might be at a different point, might have a 
different value, on each of the three gradients. Assume now that 
response E, which is highest on gradient 1, is at the intersection of 
three more gradients and that one of these gradients is at the inter- 
section of an additional three gradients, and so on ad infinitum. It 
is clear that we have a network in which the reinforcement or extinc- 
tion of any response (thought) whatever has the possibility of re- 
inforcing or extinguishing any response which is on a gradient in the 
network, provided that the sequence of responses leads to responses 
which are at the intersection of gradients. 

It is not clear from Bollard's and Miller's discussion whether 
thoughts as responses are considered preparations for action. How- 
ever, Sperry [6], on the basis of neuroanatomical considerations, so 
considers them, and such a view seems consistent with their theory. 
Considering thoughts as preparatory responses, then, it follows that a 
train of thought is a resultant of approach-avoidance gradients, and 
that which thought follows which depends on the resolution of con- 
flict at the intersection of generalization gradients. The resolution of 
conflict at any point in the network would then depend upon the 
heights and slopes of approach-avoidance gradients to the number n, 
where n is the number of intersecting generalization dimensions. Pre- 
dicting the next thought (response) at an intersection where n = 10 
would obviously present certain complexities of analysis and prob- 
ably could best be handled by considering the network as a whole in 
terms of a probability model. 

If, in a general sense, the ego is considered the "executive of the 
personality," then the ego would, in the terms here used, be the larg- 
est network of response probabilities. I say largest because I am not 
neglecting the possibility of "split" personalities. In my opinion, 
this definition conforms to the clinical observation of unity in per- 
sonality. The network is a system of response probabilities. It meets 
the notion that the ego is the conflict area of personality, and it meets 
the concept of "centrality" or of "consistency" in personality. The 
responses at the intersections of the gradients are certainly "central" 
in the sense that reinforcement or extinction, avoidance-of and ap- 
proach-to the intersections, will have the most widespread and diffuse 

124 Kentucky Symposium 

The response-cue character of thoughts and the notion of the col- 
lapse of stimulus and response generalization dimensions and of 
response sequences into single gradients or dimensions are also con- 
sistent with the concept of self. In terms of social learning and the 
course of socialization it seems that the most "central" thoughts, those 
at the intersection of the most gradients, would be self-concepts, 
thoughts which prepare the individual to respond to himself in given 
ways rather than others. The reinforcement and extinction of these 
thoughts would have the most widespread effects. Thus we might 
define such central thoughts or self -responses as constituting the "self- 
concept." These notions could be carried further, but I am not de- 
veloping a theory of the ego and of the self; I am merely trying to 
indicate that, as is the case with psychotherapeutic propositions, the 
notions of therapy-based personality theories may also be translated 
to a single frame of reference. 

The implications for psychotherapy of such a formulation seem to 
be obvious. If a person is neurotic or maladjusted, then these "central" 
thoughts (self-concepts) are, or tend to be, the cues for anxiety re- 
sponses. Then all other thoughts connected with the central thoughts 
also become cues for anxiety, and thoughts which are quite different 
from the central thoughts and are on generalization dimensions re- 
moved from the central thoughts may come to evoke anxiety. And 
similarly for extinction. 

Suppose now that anxiety is reduced by a typical instrumental be- 
havior. Then all thoughts producing the anxiety through the gener- 
alization network become cues for the anxiety-reducing behavior, pro- 
vided they are close enough temporally to the anxiety-reducing be- 
havior. If the central thoughts are extinguished in ways which are 
described by Dollard and Miller, then all thoughts connected with 
them by generalization gradients may receive their increment of ex- 
tinction, and hence the probability that the instrumental behavior will 
be evoked should be lowered. 

Now we add the clinical observation that the most threatening 
behaviors tend to be those which call into question the "traits" and 
"roles" of an individual. Taking this observation together with the pre- 
ceding development, we can perhaps glimpse how a theory of the ego 
or self would have implications for psychotherapy. 

One implication is that "free association" is not at all free and that 
it tends to lead directly to central (self) thoughts; also that, if central 
thoughts are cues for anxiety, the procession of the response sequence 
arouses more and more anxiety when the central thoughts are anxiety 

Butler 125 

producing with the consequence that avoidance gradients are then 
found to be above approach gradients. This is in agreement with 
nearly all theories of therapy. 

Another implication is that a client may get closer to his central 
thoughts from the "periphery" of the network if he is in a therapeutic 
situation in which the external cues arousing anxiety are minimized; 
i.e., the tolerance of the client for self-evoked anxiety cues is maxi- 
mized. It follows that more central thoughts (emotional problems) 
of the client can be "approached" before anxiety-reducing behavior 
(avoidance) is evoked. 

This result is the same as the one reached earlier on a more limited 
basis; it is also more analytical. 

Another implication is that, if the ego or self is a network, then the 
spread of reinforcement and extinction which occurs obviates the 
necessity of working through all central problems. Lowering of the 
cue value of thoughts for anxiety means that "internal courses of 
action" are not now as threatening as they were, and, therefore, that 
different courses of instrumental behavior (social interaction) will 
eventuate. In another paper \1] I advanced the notion that courses 
of instrumental, interpersonal behaviors, since they are largely antici- 
patory, tend to evoke consequences (response- wise) which reinforce 
that behavior. In the context of this paper, it is clear that the se- 
quence of internal cue-producing responses is part of the se- 
quence of instrumental behavior and therefore should be reinforced 
along with it, providing a double source of reinforcement for the 
internal responses. This explains why therapy can have such dra- 
matic results; the anticipatory behavior of the client is influential in 
creating his own social (response) environment. 

Finally, from the perspective that learning theory can give us, it 
seems that all this implies that we should look more and more closely 
at the role of the therapist in evoking and reducing anxiety reactions. 
If we do this with an open mind, I think that some of the contradic- 
tions, inevitabilities, and absolutes implied by the existence of so 
many psychotherapeutic systems will seem not to be so contradictory, 
inevitable, and absolute. 

With respect to the contributions of the domain of learning to re- 
search in psychotherapy, it is my conviction that at present learning 
theory as such is able to contribute but little to actual research on 
the process of therapy. It seems to me that this is so because, as I 
stated earlier, translations or reductions of therapeutic theory to learn- 
ing theory will not, in general, prove anything save that the sets of 

126 Kentucky Symposium 

propositions are consistent or inconsistent, and the question of the 
actual denotata of the therapeutic propositions is not solved by the 
process of translation, although the translations may be richly sugges- 
tive. As far as I can see, the best use of the domain of learning can 
come from applying the experimental techniques invented in the 
laboratory to tap processes and abilities which are used in everyday 
life and thus may be substituted for the usual measures of social ad- 
justment which are so unsatisfactory. Let me give an example of 
the type of research that may be used in studying psychotherapy, a 
type of research that to my knowledge has not yet been done. 

It will be assumed first that maladjusted individuals may be di- 
vided into two classes, "repressive" and "vigilant." The repressive 
person tends to withdraw from threatening stimuli by "not noticing" 
them, by "forgetting" them, etc. The vigilant individual tends to 
"enter the stimulus field," to deal with the stimuli. He does not for- 
get them. He keeps them in awareness, intellectualizes about them, 
and can and does talk about them. Lazarus et al. [4] have presented 
experimental evidence for the existence of such types and have shown 
that they can be identified clinically. 

Assuming that our subjects, who are to undergo psychotherapy, 
have been identified, the following procedures are followed: 

1. An association test, containing trait names or attributes drawn 
from psychotherapeutic protocols, is administered and scored for 
complex indicators. The hypothesis is that vigilant individuals will 
be differentiated from repressive by the exhibition of some extremely 
fast reaction times. 

2. The words in the association test are administered tachistoscopi- 
cally at speeds too high for correct recognition, with galvanic skin re- 
sponse and reaction time being used as criteria of subception. The 
hypothesis is that traumatic or anxiety-producing words can be identi- 
fied through this procedure. 

Clinical assessment, complex indicators, and subception are to be 
used to stratify the experimental population. 

3. The words in the association test are then sorted by the subjects 
along subjective metrics of self and ideal. The hypothesis is that 
words with high scale discrepancies are consciously threatening. This 
has already been shown experimentally by Haigh [3]. 

Now, taking words which simultaneously (a) are complex indi- 
cators, (b) are associated with subception, (c) show high self-ideal 
scale discrepancies, we construct the following learning tasks: (1) 

Butler 127 

word mazes, inserting the "threatening" words at points favorable 
for recall and having the subjects learn the maze to a given criterion; 
(2) paired-associate lists, using nonsense syllables for the first lists 
and judicious admixtures of neutral and threatening words for the 
second lists. 

Finally, repeat the above procedures after therapy. 

The hypotheses are as follows: 

1. Repressive individuals will forget the threatening words and 
remember the neutral words. 

2. Vigilant individuals will forget neutral words close to the threat- 
ening words and will remember the threatening words. 

3. After therapy both groups will be more efficient with respect to 
acquisition and retention. Of course suitable controls and correc- 
tions must be made for initial learning. 

4. After therapy both groups will exhibit fewer threat reactions to 
a list comparable to the pretherapy list. 

Such procedures, drawing upon already established experimental 
results, could be expected to show whether therapy had an effect 
upon tasks involving stimuli shown from the domains of learning, per- 
ception, and personality theory to be related to the conflicts of the 
individual. It should be noted also that such tasks have a certain 
resemblance to everyday learning tasks which require serial learning, 
associative learning, and accuracy of perception under conditions of 

To sum up, it has been my purpose in this discussion to indicate, 
from a clinician's standpoint, contributions to the theory and prac- 
tice of psychotherapy that might accrue from the translation of psy- 
chotherapeutic propositions to those of learning theory. The main 
advantage theorywise appears to be the possible reduction of disjoint 
propositions to a common frame of reference to the effect that new 
possibilities may appear and be explored systematically. The main 
contribution of the techniques of investigation involved in the learn- 
ing laboratory seems to be confined at present to the objective study 
of the outcomes of psychotherapy. 


1. Butler, John M., The interaction of client and therapist, J. dbnorm. soc. Psy- 

chol, 1952, 47, 366-378. 

2. Dollard, John, and Miller, Neal E., Personality and psychotherapy, New York, 

McGraw-Hill, 1950. 

128 Kentucky Symposium 

3. Haigh, Gerard V., The role of value and threat in perceptual orientation, un- 

published Ph.D. dissertation, University of Chicago, 1951. 

4. Lazarus, Richard S., Shaffer, G. Wilson, Fonda, Charles P., and Heistad, 

Gordon T., Clinical dynamics and auditory perception, Amer. Psychologist, 
1950, 5, 305-306 (abstract). 

5. Mowrer, O. H., The law of effect and ego psychology, Psychol. Rev., 1946, 53, 


6. Sperry, Roger W., The mind-brain problem, Amer. Sci., 1952, 40, 291-312. 

Learning: an Aspect of 
Personality Development 


I must confess that when I came here I was worried about how some 
of the things I have to say would be received. When I began to con- 
sider the topic of this symposium some of the things I thought of 
seemed so extreme and so bad tempered even to me that I wondered 
if they could really be true. 1 am accordingly very grateful to the 
other participants for the reassurance I have received from their 
remarks. I am particularly grateful to Dr. Spence for his disarmingly 
frank and modest statement that his learning theory at the present 
time has no application to any but the most simple behavior. The 
other participants seem to be in substantial agreement, not only about 
Dr. Spence's theory but also about learning theories in general. 

I suppose 1 should be happy about this situation because it re- 
moves much of what I have to say from the realm of controversy 
and consequently from the risk of emotion and hard feelings, but as a 
psychologist I find it somewhat depressing. 

From any practical point of view the basic problem of psychology 
is the problem of learning. Most psychologists are paid to help people 
to learn. Whether we are academic psychologists, educational psychol- 
ogists, clinical psychologists, or industrial psychologists we are supposed 
to be experts in learning. This is embarrassing because the truth is that 
nobody knows very much about learning. Hilgard, writing in 1948, 
sixty-two years after Ebbinghaus, said, "There are no laws of learning 
that can be taught with confidence. Even the more obvious facts . . . 
are matters of theoretical dispute" [3, p. 326], In psychology theories 
of learning are taught and argued about but not used. The truth is 
that the main advantage the practicing psychologist has over an in- 
telligent layman in dealing with learning is his greater experience. 

The inadequacy of our current learning theories for practical pur- 
poses is clearly revealed by their failure to have any effect on educa- 
tional practices and objectives. Only a few years ago the psychology 


130 Kentucky Symposium 

of learning constituted the core of educational psychology. Now 
learning has been pushed far into the background by mental hygiene 
and personality. Although a few teachers try to give themselves an 
air of authority by talking about conditioning when they mean learn- 
ing, the practices of the typical school have remained completely un- 
affected by any learning theory of the last 35 years. 

This seems strange when we remember that learning first became 
an object of psychological research because of the practical needs of 
teachers. As a psychological concept, learning is rather new. Bald- 
win's Dictionary of psychology, published in 1902, did not even list 
the term, and Warren's 1934 dictionary referred it to applied psy- 
chology only [1], Learning, as English has pointed out [I], is an 
educational concept. It was no accident that compulsory education 
laws were soon followed by Thorndike. Confronted for the first time 
with large numbers of pupils who were not prepared to learn what 
the teachers were prepared to teach, educators turned to psychology 
for help. 

Many of them are still looking. The rest have given up. 

The sad truth is that, after 50 years of careful and honest and 
occasionally brilliant research on the nature of learning, the only 
people who can be proved to have received any practical benefits 
from learning theory are the learning theorists themselves. The very 
inconclusiveness and complicated nature of our current learning the- 
ories, which make them useless to applied workers, have proved to 
be occupational assets to the learning specialists. They can, if they 
wish, make rather good professional careers out of attacking the weak 
points in one another's theories, much like the shipwrecked Scotsmen 
who made a good living by taking in one another's washing. 

My gibe at the complicated nature of current learning theory may 
seem unjust. It is not the fault of the theorists, some may feel, if 
their theories become too complicated for teachers to understand or 
for applied psychologists to use. I am not so sure. Teachers are not 
particularly stupid, and most really productive ideas in the history of 
science have been fairly simple. Many theories are complicated be- 
cause a great many qualifications have had to be added to patch up 
an idea or a conceptual scheme which was not very good in the first 
place. There are clear signs that our psychologies of learning have 
reached this futile stage of patching what had better be thrown away. 
Harlow [2, p. 27] has had something to say about this. Writing about 
learning theory, he comments, "A strong case can be made for the 
proposition that the importance of the psychological problems studied 

Snygg 131 

during the last 15 years has decreased as a negatively accelerated 
function approaching an asymptote of complete indifference." 

Of course, I may be wrong. It is just possible that one of our 
presently inadequate theories may, in the future, be developed into 
something really effective. But I don't think so. The kind of gen- 
eralization that most learning theorists are using is not the kind of 
generalization which can be applied to the behavior of actual people 
in social situations. Somewhere our analysis of learning has taken a 
wrong turn. I am inclined to agree with English [1] that the mis- 
take was made when it was first assumed that learning is a relatively 
independent psychological process. 

There are very good reasons for questioning the usefulness of the 
"process" concept of psychology itself, but Krech has protested so 
effectively [4] against this particular way of fragmenting psychology 
and people that I do not need to burden you with my version. The 
important thing is that, as English pointed out in his 1951 address 
to the Division of Educational Psychology [1], the phenomena which 
we are accustomed to group in the category of "learning" phenomena 
are lumped together for practical reasons, not for systematic or the- 
oretical reasons. Learning is always defined in terms of practical 
results. To prove learning we have to prove increased efficiency in 

"Before psychology took over, the verb c to learn' was used in com- 
mon speech to designate the whole complex process of reorganizing, 
improving, adjusting one's behavior" [1, p. 328]. When psychologists 
took up the study of learning they had no safe grounds for making 
either one of two assumptions that underlie all theories about the 
"learning process." 

The first assumption is that all learning, i.e., all improvements in 
behavior which cannot be ascribed to maturation, is the result of a 
single psychological process. The second assumption is that this 
hypothetical process is not the cause, except indirectly, of any other 
changes in behavior. If either one of these assumptions is not true 
the postulate that learning is a single psychological process cannot be 

Actually, anyone who tries to deal with learning in relation to other 
aspects of human behavior must eventually give up one or the other 
of these assumptions. Psychologists who begin by equating learning 
with the process of association are bound, sooner or later, to run into 
"learning" behavior which does not fit the association pattern and to 
conclude that learning is not one process but two or more. Those of 

132 Kentucky Symposium 

us, on the other hand, who take an open dynamic field as the model 
arena for learning behavior are sure to find the model adequate for 
dealing with other aspects of behavior and to conclude that learning 
is not an independent process at all but simply another manifestation 
of field organization. After contending for some years that learning 
is not two processes but one I have just come to realize, width the help 
of a few nudges from English [I], that what I have really meant is 
that learning is not a separate or unique psychological process at all. 

Be that as it may, the assumption that learning is a more or less 
independent psychological process has had a definite and, I believe, 
unfortunate effect on learning research and theory. 

In the first place, if we believe that learning is a process, it is quite 
natural for us to begin to think of learning as somehow independent 
of the people who learn. The time is past when psychologists were 
looking for the "true" form of the learning curve. But it is still 
standard practice to obscure the individual nature of behavior and 
learning by combining the data from several subjects. 

In fact, if learning is a process it is not necessary to study people 
at all. White rats are much more prolific than people, are cheaper to 
feed and house, and are usually willing to work for room and board. 
Do not misunderstand me. I like white rats. When properly handled 
they are gentle and often affectionate. Some white rats are better 
company than some people. I have a high respect for their intelli- 
gence. But I should like to say that, whereas all psychologists are 
properly aware of the danger of assuming that rats are just like people, 
many of them do not seem sufficiently aware of the dangers involved 
in assuming that people are just like rats. 

If we assume that learning is a process and that it is the task of the 
learning psychologist to study that process it docs seem to follow 
that the place to study it is under the simplest possible conditions, 
probably best represented by a white rat in a single section T-maze. 
No matter what apparatus is used it is sure to be one which limits the 
possible behavior of the subject to two simple alternatives at each 
choice point. This is done to simplify and facilitate record keeping. 
Then, so that the animal will not learn too fast, all but one of the 
possible cues that it might use for solving the problem are eliminated. 
Personality differences are further minimized by arranging that all 
subjects share a common physiological tension, usually hunger, during 
the experiment. 

Such practices would make no sense if the experimenters were 
trying to learn about people or even about rats. But they follow 

Snygg 133 

quite logically if the experimenter is trying to study a process. Since 
the behavior of the subject in most experiments is pretty well limited 
by the nature of the apparatus that is used, what we may actually 
be doing is studying the limitations of our apparatus. 

At any rate we have reached something that looks like a dead end. 
It may not be. I have been wrong before and expect to be again. 
But just in case it is a dead end we should be calling for volunteers 
to go back and try to find another approach. 

Anyone who tries to do this will be in new territory and will have 
to go where there are no maps. In a science as new as psychology 
we cannot afford to overlook any possibilities, and I hope that a num- 
ber of possible approaches will be explored. But if I had to take 
my choice I would go back and start where the early "process" the- 
orists didwith the practical problems of the people who are hired 
to help others to behave more effectively. 

If we do so we shall find that this group and its problems have 
changed a great deal since learning psychology left it and went to the 
cats, rats, and dogs 50 years ago. For one thing the group now 
includes therapists as well as teachers; for another, the teachers are 
dealing with problems that are different from those that engaged them 
then. Teachers are now much more aware of their responsibility for 
the development of citizenship, character, and effective personality 
than they were 50 years ago, when it was generally assumed that 
whoever was well informed would also be good. 

It is too bad that psychologists became interested in learning before 
this happened. By setting up our definition of the learning "proc- 
ess" too soon we have cut the learning psychologist off from the 
major problem of our schools and left it in the exclusive possession of 
the mental hygienist. And by making a dichotomy between the learn- 
ing of facts and the development of personal character and attitudes 
we have encouraged the teacher to make a similar distinction and to 
concentrate on one or the other. It was several years ago that a 
teacher told me that character education was awfully important and 
that she intended to do something about it sometime. She would, 
too, if she could just get all her pupils through the multiplication 
table in time to do it. But I think it could happen today. 

Now how can we repair this fracture and get the changes in an 
individual's general style of behavior which we have called person- 
ality development related to the changes in his specific acts which we 
have called learning? Stated in that way, there is an obvious rela- 
tionship. If you are a behaviorist the acts are apt to be considered the 

134 Kentucky Symposium 

result of the formation of separate S-R connections, and personality 
change is visualized as being nothing more than the sum total of 
these individual learnings, each one of which is essentially unrelated 
to the others. 

I don't like this because it does not seem to give sufficient weight 
to the fact that in a situation where an individual is free,to choose 
among a number of alternatives his behavior will show a marked de- 
gree of individuality and consistency. His learnings will conform to 
his personality pattern at the same time that they are changing it. 

The following report, written by a college student, is an example 
of the sort of thing I am talking about: 

During my years in grade school 1 was always at the top of the class. 
I always got very high grades in all subjects but especially in arithmetic 
and silent reading. I spent a good deal of school time reading encyclo- 
pedias for fun and scored 100 in the eighth-grade arithmetic 1 Regents' 
examination while I was still in the seventh grade. 1 also won the county 
spelling contest that year. By passing the Regents' exams during my sev- 
enth year I was ready to go to high school, but during the summer I sud- 
denly decided that 1 didn't want to go to high school and my parents 
couldn't understand the reason. 

It was simple. My parents were strict members of a very strict religious 
sect. I would rather face a firing squad at dawn than faee my father ii 1 
had "sinned." That summer my mother somehow found out that I was 
masturbating. She pointed out the village idiot and explained that she 
got that way by masturbating. She also threatened to tell my father. 

I decided that I had ruined my brain and could never learn anything 
more. I was afraid that 1 would flunk high school courses and my father 
would find out the horrible truth. I started to school with that idea in 
mind and barely scraped through the first year with the exception of math. 
I flunked that twice. 

At the beginning of the second year an intelligence test was administered. 
The students were riot allowed to know the results, but a boy who was 
working in the NYA program told me that I had a score of 90 which was 
about average. My work began to improve a little. Later that year one 
of the boys got a book on sex, and "eminent authorities" stated that mas- 
turbation did not lead to insanity or mental deterioration. I graduated 
from high school fourth in the class. 

When I entered the army I scored 145 on the Army General Classifica- 
tion test, and in three verbal intelligence tests since then I have con- 
sistently scored high (not under 135). 

For two years the idea that I was stupid and had ruined my brain kept 
my school grades low and even depressed my IQ thirty to forty points 
below my usual score. 

If we think of learning theory and personality theory as independent 
areas of psychology we are at a loss when we have to deal with sit- 

Snygg 135 

nations like this. Learning theories are set up to explain why people 
change, and personality theories are set up to explain why they don't 
change. As a result personality theory is incompatible with learn- 
ing theory, and we cannot work out any relationship between them. 

When we are watching an instru mentally depersonalized animal 
in a T-maze or problem box this is not likely to seem important. But 
when the task is to predict the behavior of an individual who is free 
to pursue any of a number of alternative courses of action it becomes 
obvious that neither type of theory is adequate by itself. And since 
they are based on conflicting premises and lead to conflicting conclu- 
sions we can't use them together. 

The conceptual scheme which I think would avoid this compart- 
mentalization has been described elsewhere [5, 6], and I do not have 
the time to describe it in detail here. But if you can remember the 
boy who "ruined his brain" I should like to call your attention to three 

The first is the catastrophic effect that one item of information or, 
rather, misinformation, had on the child's personality pattern. This 
personality change is quite obviously not the result of the mere ac- 
cretion of one more fact. 

The second is that the kind of learning of which he was thereafter 
capable was limited and determined by his view of the situation. As 
a good student he had excelled in arithmetic; as a boy who had ruined 
his brain he failed algebra twice. 

The third is the way later items of information acted as precipitat- 
ing agents to throw the personality into a third pattern in which he 
was again an able student. 

Here we have the main characteristics of a dynamic field operat- 
ing on something like all-or-none principles with change by trans- 
formation rather than by addition. The parts of the field are so 
interdependent that they are very resistant to change. But when 
change is possible and does occur all parts of the field are affected. 
If we adopt a dynamic field as the model for our conceptual system 
it is easy to avoid the separation between learning theory and per- 
sonality theory that has caused so much trouble. Let us, for instance, 
assume that all behavior is determined by the behaver's perceptual 
field at the instant of action. This is a field theory, since the per- 
ceptual field is an organized field with space-time dimensions which, 
like all dynamic fields, tends to remain organized in the face of ex- 
ternal interference. From this point of view, the consistency, related- 
ness, and stability of an individual's behavior, which constitute the 

136 Kentucky Symposium 

problem of personality psychology, are the natural results of the or- 
ganized nature of the causal field. An individual shows a particular 
style of behavior as long as his perceptual field maintains a particular 
pattern of organization. At the same time all the specific acts of the 
individual, changes in which constitute the problem area of learning 
psychology, are simply the ways in which the organization and in- 
tegrity of the perceptual field are maintained. We avoid pain, move 
from the hot sun to the shade, seek food when we are hungry, and 
rest when we are tired to maintain the balance between the self and 
not-self in the perceptual field. These acts could also be considered 
as resulting from aspects of homeostasis in a physiological field or as 
physiological drives. 

But we do other things which cannot be considered to have homeo- 
static or physiological value without seriously changing the meaning 
of those terms. I am referring to such things as martyrdom and 
suicide and writing books when no profit is expected and the actions 
of Harlow's monkey in putting in a 19-hour day keeping a window 
shade up. The perceptual field, as an organized entity, includes a 
perception of the future, at least of its existence. Maintenance of a 
satisfactory relationship between the perceptual self and this vaguely 
perceived future requires unceasing vigilance and activity. The fu- 
ture is uncertain. Since it is uncertain we can never gain a perma- 
nently satisfactory relationship with that part of the field. Under the 
threat of the future we can maintain the organization of the percep- 
tual field only by striving for better and better relations between our 
perceptual selves and the rest of the universe through an increased 
feeling of worth, value, power, and acceptability. 

From this point of view the primary goal of all behavior is the 
achievement of a more adequate self. Personality development is 
the goal of all our acts, and learning is the result of our attempts to 
achieve it. 

Can this theory, which does not give any particular priority to the 
so-called physiological drives, be applied to the behavior of rats? I 
think so. Although a rat's concept of the future is probably even 
vaguer than our own, any animal with distance perception can be 
assumed to have some perception of time [6, p. 350]. Many of the 
latent learning experiments have shown learning without apparent 
tissue-tension motives, and I have seen starving rats give up food 
and water in order to keep possession of an activity wheel from their 
cage mates. The theory certainly fits the behavior of Harlow's mon- 
keys [2]. 

Snygg 137 

It seems to me that this conceptual scheme has a number of ad- 
vantages. On a theoretical basis it provides a much better frame- 
work for predicting behavior in a free situation than any of the 
specialized learning theories because the personality of the learner, 
in the form of his perceptual field, is an integral part of the conceptual 
scheme. If our investigations of learning behavior are to be helpful 
to the teachers, psychologists, and other people who are interested in 
assisting personal development, we will have to stop thinking of 
learning as an independent process and shift to some kind of unified- 
field theory. The one I have sketched happens to be the one I like. 

Whether you like that particular scheme or not, I hope that you 
will consider the disadvantages of the present isolation of learning 
theory which I have tried to point out. Whatever conceptual scheme 
we use, psychology, education, and psychiatry could all benefit from 
a greater use of human subjects in research, with animal subjects 
being reserved primarily to check the validity of hypotheses de- 
veloped from the observation of human subjects in situations where 
there are enough courses of action to give individual patterns and 
idiosyncrasies a chance to show. 

If we do this I am confident that we shall find that the learner is 
not the passive victim of his environment but an active explorer and 
creator of his own world. He is not a puppet at the mercy of the 
stimuli which bombard him or even of his own hunger pangs. In 
perceiving, in learning, in forgetting, in imagining, in rationalizing, 
he is selecting from all the potential aspects of his world those which 
best satisfy his need for personal growth and development. If we 
forget the active and purposive role which the learner plays in the 
achievement of his own personality and his own future we are bound 
to deal with him arbitrarily and unrealistically and to frustrate him 
and ourselves. 


1. English, H. B., Learning "they ain't no such animal," J. educ. PsychoL, 1952, 

43, 321-330. 

2. Harlow, H. F., Mice, monkeys, men, and motives, PsychoL Rev., 1953, 60, 


3. Hilgard, E. R., Theories of learning, New York, Appleton-Century-Crofts, 


4. Krech, D., Notes toward a psychological theory, J. Pers., 1949, 18, 66-87. 

5. Snygg, D., The psychological basis of human values, in Ward, A. D. (Ed.), 

Goals of economic life, New York, Harper and Brothers, 1953. 

6. Snygg, D., and Combs, A. W., Individual behavior, New York, Harper and 

Brothers, 1949. 


Theory and Measurement 


I should like to start off by indicating briefly what I hope to cover in 
this paper. The major sections will deal with the following topics: 

First, "theoretical" versus "antitheoretical" approaches to the un- 
derstanding of behavior. This section concerns itself with a sugges- 
tion as to a distinction in the use of the word "theory/' and with an 
indirect analysis of why people are often opposed to rigorous theoriz- 
ing about the behavior of organisms. 

Second, part of a theory or postulate system dealing with "errors" 
The phenomena with which the theory deals are some of those men- 
tioned by Freud [2] in The psychopathology of everyday life, while 
most of the basic concepts are derived from learning theory. The 
"complete" theory will not be presented here because of time restric- 

Third, a brief report of an exploratory experiment designed to test 
roughly one of the deductions from the theory. In this experiment 
students, taking a final examination of the essay type, were given rela- 
tively hard or relatively easy questions. "Errors" were measured in 
terms of number and type of writing deviations. 

Fourth and finally, some suggestions as to situations lending them- 
selves to manipulations of the type necessary to test and extend the 

What about the value of theories in psychology? Are they per- 
forming an intellectual service, or are they premature and otherwise 
undesirable? In the last few years there has been a series of attacks 
on theories and theory construction in psychology. It is my impres- 
sion that by far the greatest number of comments on rigorous theoriz- 
ing are negative. For this reason, I was particularly pleased at 
receiving an invitation to participate in this symposium, and to have 
an opportunity to discuss this apparent trend. I should like to raise 
a small voice against the antitheoretical points of view. 


Ammons 139 

First, I should like to suggest a useful convention in talking about 
theory. When I use the term "theory/' I am referring to a rigorous 
postulate system, the simplest form of which is the Aristotelian syl- 
logism. Any formulation short of this rigor, I prefer to call specula- 
tion, without, however, in any way reflecting on the potential validity 
or utility of the ideas. Thus Guthrie's and Tolman's learning theories 
would not be called theories, but complex speculation about possible 
theorizing in psychology. E. J. Gibson's postulate system [3] dealing 
with the effects of generalization and differentiation on verbal learn- 
ing would be called a theory. Hull's italicized statements in Prin- 
ciples of behavior come close to being theory in this sense. 

This convention is useful and, I believe, should be adhered to for 
its psychological effect. I imagine that each of us here has had the 
experience of speculating about a phenomenon, and feeling that his 
speculations account for it quite adequately, only to find that when 
these speculations were reduced in writing to a minimum set of 
statements there was some "fatal" inherent contradiction or glaring 
hole in the reasoning. By carefully distinguishing between the loose 
and the logically precise system, between speculation and theory, one 
reminds oneself of the long step between them. The ad hoc loose 
system such as that of the early Gestalters [5] is capable of predicting 
everything or nothing, whereas the phenomena predicted by the 
Mathetnatico-deductive theory of rote learning [4] are usually criti- 
cally testable. 

I recently had the interesting experience of attempting to teach a 
highly selected group of graduate students how to go about formulat- 
ing simple theories. In fact, we spent a whole N semester working on 
this problem. As a clinician, I was fascinated by the responses of 
these students to the idea of constructing theories and tried to make 
observations as to the possible bases of these responses. Let us make 
the assumption that the present widespread attack on theories and 
theorizing in psychology is merely the organized expression of atti- 
tudes and feelings widespread in our culture. Then by observing my 
students I would be able to come to some sort of tentative conclusions 
about the reasons for the attack. 

What difficulties did the students have, and what were their ob- 
jections to rigorous theorizing? 

The first problem was comprehension. A postulate system is hard 
for most students to understand. They say, "Why not put it in simple 
language?" Now, an examination of this comment shows that it is 
somehow at variance with the facts. The simplest, most concise way 

140 Kentucky Symposium 

to express a complex set of relationships in words is to set up a pos- 
tulate system. Any attempt to substitute literary language and ex- 
position for the postulate system leads to the use of large numbers of 
words and a great decrease in precision, especially in regard to im- 
plications. Why, then, is the system hard to understand? My stu- 
dents discussed this point at considerable length and concluded that 
the difficulty for them was the result of inadequate training, particu- 
larly in mathematics. They felt that they simply had not been taught 
to think clearly and simply. This is all very well, of course, but how 
are we to persuade students to take mathematics? How many of us 
who serve as advisors are willing to "go back" and make up our own 
deficiencies in logic and mathematics? The average amount of train- 
ing of psychologists in symbolic logic and mathematics is, I am afraid, 
appallingly small. 

The second objection to postulate systems was that they call for 
operational definitions of terms, and when we define terms this way 
we leave out meanings. That is, the words lose some of their glorious 
essence. I call this a "something-more-than" attitude and have pointed 
out to persistent objectors along this line that they might too be called 
"something-more-thaners." This attitude grows from a satisfaction 
with things as they are and a failure to realize the nature of language. 
I have often pointed out that the meanings of terms are not inherent, 
but are given by experience. By setting up methods for more sharply 
defining words we are bound to step on somebody's toes. Interest- 
ingly enough, it is most often the private, emotionalized meaning 
which is left out in this process, and which the "something-more- 
thaner" tries to force us to include. 

A third, and related, objection to logical systems is that they are 
incomplete and artificial. They leave out much that is obviously 
present in the world, and lead to the ignoring of important facets of 
the phenomena. Here again is the "something-more-than" feeling 
rearing its ugly head. I like to point out to the student at this point 
that there is nothing to keep him from expanding the system to in- 
clude those things which he feels have been neglected. He often 
counters that they could never all be put in, so the system cannot have 
any great value and may even mislead by oversimplification. I have 
found that careful investigation of the student's feelings usually 
reveals a real resistance to thinking carefully and precisely about 
the phenomena because of some kind of often hidden, personal needs. 
That is, he has some peculiar personal reservations about dealing with 
this sort of thing coldly and intellectually. 

Ammons 141 

The fourth objection is in some ways similar to the third. It takes 
the form, "We don't know enough to set up a postulate system about 
that." This objection often seems to grow out of a desire to avoid 
the rigors of careful logical thinking. The student would prefer to 
go on carrying out relatively unrelated experiments to obtain em- 
pirical results ad infinitum. This is the Baconian approach, of course, 
and perhaps grows from a need to hoard relatively unorganized in- 
formation. To this objection, I usually point out that we never will 
know everything about anything and that the outcome of adopting 
this approach might well be a permanent avoidance of theorizing. 
Several students have then pointed out that "premature" theorizing, 
or theorizing before we have "enough" data, may lead to a blindfold- 
ing of future research. I am quite willing to agree that this has 
happened and probably will continue to happen. However, the per- 
son who allows a theory to restrict his thinking, or who becomes 
violently partisan in an attempt to "prove" the theory, would prob- 
ably be just as dangerous if there were no theory. He would then 
probably feel compelled to defend some particular finding or findings 
in the same sort of blind partisan way. But let's not blame the theory 
for the scientist's personal shortcomings. 

In general, there are some obvious reasons why the student, and 
for that matter almost anyone, might and often does feel antitheoreti- 
cal. Constructing or even comprehending a "simple" postulate sys- 
tem is hard work and is often not immediately rewarding. Then, 
when we have a system worked out, it ordinarily calls for continuous 
revision and extension more work. For the system to be of real 
value, there must be a great deal of systematic work done, and we 
Americans seem to find it hard to work systematically for long periods 
of time. Our individualism also often keeps us from working on 
problems which have the misfortune to have been suggested by some- 
one else. Finally, there is a real anti-intellectualist feeling on the 
part of many "clinicians" who, probably rightly, find an intuitive ap- 
proach most valuable in practicing their art, and then generalize this 
into opposition to "sterile, meaningless theorizing." 

Well, I might say that trying to teach my class something about 
theorizing was a stimulating experience, but I uncovered no particular 
talent for theorizing. 

Let us turn now to the principal topics of this paper, a tentative 
theory of errors, a report of some related research, and some proposals 
as to how to study the situation experimentally. I should like to start 
out by defining in a tentative way the terms used in the theory. 

142 Kentucky Symposium 

Error: A response other than that appropriate to the motor set 
present, where this response is appropriate to other parts of the 
stimulus complex. 

Response: Observable striated muscular behavior by the individual. 

Motor set: Bodily orientation for the performance of a given be- 
havior, inferred jointly from the instructions given by the experi- 
menter or subject to himself and the physical orientation of the person. 
We can to some extent get at it by asking the subject what he intends 
or intended to do, or by setting up an objective criterion for determin- 
ing whether or not the physical orientation would allow the perform- 
ance of the task. 

Appropriate response: The response which the individual says he in- 
tends or intended to make and for which he is physically oriented is 
the appropriate response to the motor set. Appropriate responses to 
other parts of the stimulus complex are those which would be most 
frequently made if those parts of the stimulus complex were dominant. 

Stimulus complex: Various components which make up the stimulus 
such as stimuli from motor set, specific drive stimuli, and external 
stimuli. Any of these can be changed relatively independently, chang- 
ing the stimulus complex. 

Dominance of a component of the stimulus complex: A drive stimu- 
lus is more dominant as the drive becomes stronger. When the sub- 
ject is asked to describe a situation, a particular stimulus component 
is dominant to the extent that it is mentioned earlier in his description. 
Frequently this dominance must be inferred from the past history of 
the individual. The report may not be accurate from the point of 
view of the experimenter, as in the case of the individual who has 
always hated a sibling and now reports that his emotion is one of love 
and affection, yet behaves as if he still hated her. This is admittedly 
shaky ground for inference in some cases. 

Drive stimuli: Those stimuli characteristically noted by the human 
organism in connection with hunger, thirst, sex frustration, fear, anx- 
iety, etc. One could infer the presence of such stimuli in terms of 
strength of drive. 

External stimuli: Environmental energies which affect the receptors 
of the organism. When the organism is oriented in such a way that 
the receptors can be affected by the energy and the energy is suffi- 
cient to stimulate the receptors, stimulation is normally assumed to 
take place. 

Strength of the response tendency: Latency of the response, physi- 
cal strength of the response, and probability of the response occurring 

Ammons 143 

in the presence of or closely following the presence of a given stimulus 

Stimulus similarity: Stimulus complexes are similar to the degree 
that they contain similar components and are relatively less separated 
along the various discriminable continua. 

Strength of drive: Might be the self-rating of the individual or 
might be inferred from the past history of the individual with respect 
to the time since drinking, time since eating, number of times a pleas- 
ant or unpleasant consequence has followed a particular stimulus 
complex, etc. Thus drive stimuli can be associated with primary or 
secondary drives as conceived of by Hull. Emotions are considered 
to be drives. 

Reward: The satisfaction of some need, goal-object consumption, 
or avoidance of noxious stimulation. 

Thus we have the basic concepts: error, response, motor set, ap- 
propriate response, stimulus complex, dominance of a component of 
the stimulus complex, drive stimuli, external stimuli, strength of re- 
sponse tendency, stimulus similarity, strength of drive, and reward. 
Although not all of these are used in the part of the theory which I 
am going to outline today, I thought it might provide a better over- 
view to include them. 

Only the first four of eleven postulates will be presented, along 
with several deductions generated by them.* It will be seen that I 
am not doing anything particularly original, just attempting to for- 
mulate some of Freud's ideas in terms of concepts from learning 
theory. Many of the approaches taken here have been suggested by 
other writers such as Mowrer [7] and Miller [6]. I consider the pres- 
ent form of the system to be tentative. 

The whole system is based upon the assumption that behavior is 
predictable. The first postulate reads as follows: 

POSTULATE 1. To any stimulus component or complex, there 
are a number of possible responses. The strengths of the re- 
sponse tendencies differ. Thus there is present a "strength" hier- 
archy of responses to any given stimulus component or complex. 

It should be noted that a response may produce stimuli for other 

POSTULATE 2. The more similar a stimulus component or com- 
plex is to another given stimulus component or complex which 

* An outline of the "complete" system can be obtained by writing to the author. 

144 Kentucky Symposium 

has regularly elicited a response in the past, the stronger the re- 
sponse of this kind now elicited by the new stimulus. 

This implies a stimulus generalization gradient. In a free associ- 
ation situation, the stimulus includes the feeling present, intellectual 
content, traces of these from previous recent verbalizations ? and pres- 
ent physical surroundings. Concepts are all related; stimuli are all 

POSTULATE 3. The stronger the drive, the stronger the response. 

This postulate will interact in effect with Postulate 2, since drive 
has a characteristic stimulus value which will be changed by in- 
creases or decreases in the drive. 

Deduction 3a. The more drive present, the less similar the ex- 
ternal stimulus need be to the stimulus which in the past has 
regularly elicited a response, for it to be elicited with the same 

In a free-association situation, a very pressing drive will "force" a 
response to relatively non-pertinent stimuli. 

POSTULATE 4. The components of a given stimulus complex 
may in isolation elicit different responses. When the components 
are combined in the stimulus complex, the greater the dominance 
of a given component and the greater the strength of a given re- 
sponse tendency associated with it, the more likely the stimulus 
complex is to elicit this response. 

Deduction 4a. If a response has been regularly elicited under 
a low drive and is now elicited with a high drive of the same 
kind present, we will observe an increase in "errors," providing 
the strongest response tendencies to the motor set and the drive 
are different and that to the motor set is dominant. 

Deduction 4b. If a response has been regularly elicited under 
one drive, and the drive is changed to another without altering 
the other stimulus components (especially motor set), there will 
be more errors, providing the appropriate dominant response to 
the drive-stimulus component from the original drive was the 
same as that to the motor set, but that to the new drive stimulus 
is different from that to the motor set, the motor set staying the 

Ammons 145 

Deduction 4c. To the extent that a single stimulus component 
dominates the total stimulus complex, the successive responses 
given by an individual will be more similar to each other. 

Strong emotion leads to stereotypy of responses, as does instruc- 
tion-induced "motor set," and the "same" physical stimulation. In free 
association, problem areas will be talked about more frequently than 
other areas. In the case of errors, we find that certain kinds are quite 
frequent, i.e., certain types of slips of the tongue and certain kinds of 
accidents in the accident-prone person. These errors should indicate 
the life areas in which the person has problems and thus be of diag- 
nostic value to the clinician. 

Deduction 4d. Other stimulus conditions being approximately 
equal, if one arouses feeling about an error, he should get real-life 
responses associated with a similar set, emotion, or drive more 
quickly than if no feeling is aroused. 

If one arouses guilt feelings, for example, by calling attention to 
an error, these will tend to become relatively dominant in the stimulus 
complex and will form a stimulus complex increasingly similar to that 
real-life situation in which the error was originally made as a response. 

So much, then, for the first few postulates and deductions. Having 
made some predictions, naturally we wanted to test one or more. A 
likely situation presented itself. The writer was teaching a class in 
history and systems of psychology, using essay-type examinations. 
Two sets of four questions were made up for the final examination. 
In both sets, questions 1 and 3 were the same, whereas 2 and 4 were 
much harder in one set than in the other, covering topics not men- 
tioned in class and only indirectly in the textbooks. Thus we had two 
relatively easy control questions for each group and two other ques- 
tions, either relatively easy or very hard. The difficulty of the ques- 
tions was checked by asking the students to rate the questions after 
the examination. Their ratings agreed with our assumptions. 

The questions were mimeographed on separate pieces of paper, one 
question to a page, and booklets made up of them, in the order 1, 2, 3, 
and 4. The two resulting sets of booklets ("easy" and "hard") were 
arranged in one random order, and passed out in this order on the 
day of the examination. Students were told that they could write on 
both sides of a page, but should write only in the booklets, and only 
during the time allotted for a particular question, which would be 
12 minutes. There was no observable difficulty in carrying out these 

146 Kentucky Symposium 

instructions. A total of 36 students completed the test and the ex- 
periment, 18 in the "easy" and 18 in the "hard" group. 

This experiment provides a rough test of Deduction 40. The domi- 
nant motor set calls for the response of writing answers clearly and 
accurately i.e., for "good" examination behavior. Students feel, dur- 
ing such examinations, numerous emotions which could le#d to con- 
tradictory behavior, feelings such as fear, anxiety, and aggression. 
The introduction of difficult questions, our questions 2 and 4 in the 
"hard" series, tends* to intensify these feelings, as indicated by the 
students' own ratings of the questions. Thus we have the dominant 
motor set and the increased drive which would lead to different be- 
havior if appropriately acted out, so that the initial conditions of the 
deduction are present. The prediction is that "errors" will increase 
differentially in the case of the "hard" questions. 

To test the prediction, we counted as "errors" all erasures, cross- 
outs, write-overs, and misspellings in each answer by each subject. 
The variance of these scores was then analyzed,* following a pro- 
cedure suggested by Block, Levine, and McNemar [1]. In effect the 
interaction between questions and groups is tested over a residual 
error term based on pooled within-groups interaction between ques- 
tions and individuals. Our "easy" group scored totals of 108, 102, 116, 
and 88, while our "hard" group scored 62, 83, 70, and 78 on the four 
questions. It will be noted that the group patterns differed in that 
the "hard" group made relatively more errors on the hard questions. 
This interaction was significant at beyond the 1 per cent level, while 
the difference between groups did not reach the 5 per cent level. 
Although the prediction and the outcome agree, one should be cau- 
tious in interpretation, since many errors can creep into experimenta- 
tion of this rather complex kind. 

We have done a good deal of thinking about other experimental 
situations which might be used in the testing of deductions from the 
system. One might ask typists in training to copy emotional and neu- 
tral selections for practice. Subjects could be asked to write in long- 
hand or shorthand about neutral and loaded topics. They might be 
asked to copy a story while studying a TAT card or other similar pic- 
torial material. Many possibilities remain to be tried out. 

To recapitulate briefly, in this paper I have tried to do several 
things: (a) to show that theorizing is a defensible procedure; (&) to 

* I particularly wish to express my appreciation to Mrs. Sylvia Post, who be- 
came much interested in the study and generously gave of her time in per- 
forming most of the statistical calculations. 

Ammons 147 

present a preliminary formulation of a theory of errors; and (c) to 
show how the variables might be measured experimentally. 


J. Block, J., Lcvine, L., and McNeinar, Q., Testing for the existence of psycho- 
metric patterns, /. abnorm. soc. Psychol., 1951, 46, 356-359. 

2. Freud, S., Psychopaihology of everyday life, in Brill, A. A. (Tr.), The basic 

writings of Sigrnund Freud, New York, The Modern Library, Random 
House, 1938. 

3. Gibson, E. J., A systematic application of the concepts of generalization and 

differentiation to verbal learning, Psychol. Rev., 1940, 47, 196-229. 

4. Hull, C. L., Hovland, C. L, Ross, R. T., Hall, M., Perkins, D. T., and Fitch, 

F. B., Mathewatico-dcdiiclive theory of rote learning, New Haven, Yale 
Uiiixcrsity Press, 1940. 

5. KofTka, K., Principles of Geslalt psychology, New York, Ilarcourt, Brace, 1935. 

6. Miller, N. E., Theory and experiment relating psychoanalytic displacement 

to stimulus-response generalization, J. abnorm. soc. Psychol., 1948, 43, 155- 

7. Mowrer, O. II., Learning theory and personality dynamics, New York, Ronald 

Press, 1950. 

Some Current Research 
Issues in Clinical Psychology 


For the most part it may be said that clinical psychologists serve in- 
dividuals in two different respects: one, diagnostic; the other, thera- 

Diagnosis comprises several functions: describing the patient, re- 
lating the patient's present condition with circumstances of the past, 
trying to understand the patient in terms of possible relationships 
between the past and the present, and, finally, applying this under- 
standing in some manner to anticipate the course of the patient's de- 
velopment. The tools for diagnosis, tests, current observations, and 
information from the case history of the patient have been empha- 
sized in our research and training; the inferential manner in which 
these tools are used in order to gain an understanding of the patient 
has received little attention. Although anticipation of the subsequent 
development of the patient requires deductive application of infer- 
ences, the rules and principles for drawing our inferences and for 
conducting our deductions are not explicit and are guided by no 
consistent theory. 

The purpose of the therapeutic process is easy to describe: therapy 
is designed to change the patient in one or more respects. Since 
many different kinds of behavioral changes are sought for patients, 
some difficulty in describing the method of therapy can be expected. 
This difficulty is compounded by the fact that there are, in effect, 
countless methods whereby clinicians seek such changes. 

Since the diagnostic function we have described can be construed 
as an attempt to understand how the patient changed from his original 
state to his present one and to anticipate how he may further change, 
and since the therapeutic function is defined as an effort to produce 
certain kinds of changes, we may say that the clinical psychologist 
is concerned with changes in human behavior. Unfortunately, as 
soon as we see the clinical psychologist as concerned with changes in 


Wittenborn 149 

human behavior, we become aware of two weaknesses:. (1) aside 
from a few esoteric laboratory aspects, our understanding of how to 
produce changes in human behavior is so obscure and incomplete as 
to be contradictory in its implications and possibly at best no better 
than "common sense"; and (2) the devices whereby we may attempt 
to predict changes in behavior do not have the benefit of empirical 
validity, and efforts to validate our tests and our practices have de- 
tracted more from our confidence than they have contributed. 

When confronted with such a suggestion of weakness, one seeks re- 
assurance and one hopes that recent research trends provide evidence 
of strength. Accordingly, some of the current literature was reviewed. 

Of the various journals that publish research reports which are 
relevant to the problems and interests of clinicians, the Journal of 
Consulting Psychology, the Journal of Abnormal and Social Psychol- 
ogy, and the Journal of Clinical Psychology are perhaps best known 
and most generally read by psychologists. The content of research 
published in these and other journals is the subject of frequent, well- 
known, and competent reviews. In our review, we were not con- 
cerned with content; we assumed that, for the most part, the content 
of the reports was relevant to the interests of clinical psychologists. 
Our review covered one year, 1952, and was designed to reveal the 
degree to which and the respects in which this research was explicitly 
or implicitly related to the problem of behavior change. An outline 
(see Table 1) was prepared which could provide a classification of 
research reports with respect to the descriptive function, the pre- 
dictive function, and the change function of the clinical psychologist. 

Parts of Table 1 may require explanation. In the Description part 
of the outline the various sources of descriptive information are classi- 
fied. For example, "Observation" as a source of descriptions may be 
informal, or it may be formal and include the use of check lists and 
rating scales. The unique feature of observation is that the observer 
participates to a relatively small degree. In sources classified as 
"Interview" the data could be recorded in any form, but there was 
an interview relationship involved. In the "History" classification the 
information descriptive of the individual's past circumstances could 
come from any prior source. It could also come from any concurrent 
source other than concurrent observations, interviews, or tests. The 
"Tests" included any standard situation for sampling the subject's 
reaction. The essential characteristic of our test source was that the 
situation was standard, relevant to some specified reactions of the 
subject, and designed to provoke reactions from the subject. 

150 Kentucky Symposium 

TABLE 1 A survey of current research for clinical psychologists 

Frequency of Various Research 
Approaches in Articles Published 
during 1952 (J. consult. PsychoL, 

J. abnorm. soc. PsychoL, and 

Classification of Research Approaches J. din. PsychoL} 

I. Description 

A. Observation 1 

1. Temporal trend inferred 

a. From past 1 

b. Toward future 

2. Concurrent material interrelated J) 

B. Interview 1 

1. Temporal trend inferred 1 
a. From past 

6. Toward future 

2. Concurrent material interrelated 3 * 

C. History 

1. Temporal trend inferred 

a. From past 

b. Toward future 

2. Concurrent material interrelated 35 * 

D. Tests (or other standard situations for 

eliciting the subject's reactions) 8 

1. Temporal trend inferred 

a. From past 2 

6. Toward future 1 

2. Concurrent material interrelated 110 

II. Prediction 

A. Stability of inferred characteristic (either 

stable traits or trends of change) 

1. Success 8 

2. Failure 1 

B. Susceptibility to change as a result of a 

specific experience (including a suc- 
cessful therapy) 

1. Success 12 

2. Failure 5 

III. Therapy change 

A. Change or attempt at change described 

only (The agent may or may not be 
specified and its effectiveness may or 
may not be anticipated) 

1. Success 9 

2. Failure 2 

B. Change or lack of change explained ex 

post facto % 

Non-research 35 

Total 266 

* Studies involving some other concurrent descriptive category also. 

Wittenborn 151 

TABLE 2 Summary of research approaches 

Per Cent of 
Principal Approaches Occurrence 

Descriptive 83 

Trend inferred (3) 

No trend inferred (80) 

Predictive (includes no purely descriptive studies) 11 

Stability tested (4) 

Susceptibility to change (including therapeutic) tested (7) 

Therapeutic change (includes no studies of differential suscepti- 
bility to change) 6 

Total 100 

NOTE: In the descriptive approaches all the historical sources and three of 
the interview sources occurred in studies which also involved test sources of 
data. Aside from this, there is almost no overlap. This is possible because 
the predictive and the therapeutic studies necessarily involve description, and 
for such studies no tally was made for description. Another kind of possibly 
overlapping count was eliminated by excluding from the therapeutic category 
all studies which involved an explicit prediction of differential susceptibility 
to change. 

Usually, when behavior is described by any of the foregoing pro- 
cedures, it is implied that the content of the description comprises a 
stable characteristic of the individual or of his behavior and that any 
change in this characteristic either is an exception or is due to some 
change in circumstances. This static emphasis is not invariable, how- 
ever. In practical clinical work clinicians frequently infer a trend 
from the available descriptive material. The trend they describe is 
usually historical, but sometimes it is conceived in such a way as to 
anticipate the future behavior of the individual. Recognition of such 
trends was sought because it could be taken as an indication that the 
research was relevant to an aspect of the behavior-change interests of 
clinical psychologists. 

The interrelation of concurrent material in some cases involved 
descriptive information from two or more of the classified sources of 
information. Almost all these overlapping cases were studies which 
involved both "History" and "Tests" information. 

The Prediction part of the outline refers to studies wherein descrip- 
tive information which was gathered at a prior time is related with 
descriptive information gathered at some subsequent time. The in- 
tervening time interval may cover a period of incarceration, a course 
of psychotherapy, or a laboratory experience. Prediction studies 

152 . Kentucky Symposium 

which examine the stability of a characteristic were in some instances 
concerned with a static trait. In other instances they were concerned 
with a growth or change process in the individual. Predictions con- 
cerning this susceptibility to change refer to all manner of reports 
wherein some prior characteristic of the individual, either naturally 
occurring or experimentally induced, was predicted to qualify the 
nature of the individual's response, i.e., change, in some other situa- 
tion. The prior characteristics of the individuals could include such 
extremes as symptoms or specific instructions by an experimenter. 
The situation in which change was differentially predicted could be 
therapy or could be a learning situation. Obviously, predictive studies 
involve description, but predictive studies were not classified under 
the descriptive heading merely because of this necessary fact. 

The Therapy or Change section does not include all the studies 
which have to do with change. Specifically, it excludes any studies 
having to do with change wherein the differential susceptibility of 
the individual to change was specifically predicted. If we except 
the susceptibility-to-change predictions, the therapy section of the 
outline refers to studies where an actual change occurred or where 
evidence of change was sought in a presumably change-inducing sit- 

As would be expected, a number of the articles did not have a re- 
search content. Their content has not been classified. Since none 
of them included any data, they could not be expected to contribute 
any new facts. 

Referring to Tables 1 and 2, one sees that our recent research has 
little concern with behavior change. Eighty per cent of the articles 
are not only independent of the topic of behavior change, but also 
they are not even suggestive of an interest in the development or the 
modification of human behavior. Four per cent of them appear to be 
explicitly concerned with the stability of behavior. Only 13% of the 
studies show a concern for describing or predicting change. Either clin- 
ical work is not much concerned with behavior change, or the research 
published by and for clinical psychologists is not much concerned 
with clinical work. 

Since an examination of our current research in clinical psychology 
fails to offer us reassurance, it may be worth our while to discuss 
briefly some of the broad areas of interest among clinical psycholo- 
gists. Perhaps such a discussion may help us to think further about 
the relevance of changes in behavior for the work and for the re- 
search of the clinical psychologists. 

Wittenborn 153 


Most frequently clinical psychologists are concerned with devia- 
tions of behavior. Such deviations are usually called abnormal. Un- 
fortunately, there are numerous ways in which abnormal behavior has 
been defined. 

Some authorities have suggested that some sort of definition of 
abnormal based on a statistical frequency is sufficient. For the most 
part these people are inclined to believe that all human attributes 
exist in varying degrees and that some degree of each of these at- 
tributes may be found in all people. They would say that, in cases 
where some particular degree of these attributes exists, the persons 
concerned may arbitrarily be classified as abnormal with respect to 
these attributes. In practice, however, such a point of view is difficult 
to employ because many of the behavioral attributes which char- 
acterize people who are regarded as abnormal or in difficulty cannot 
be confidently seen as an extension of well-known and explicitly de- 
fined continua which characterize everyone. Another difficulty in the 
statistical-frequency point of view lies in the fact that the nature of 
relevant characteristics of individuals, as well as the frequency of any 
degree of such characteristics, seemingly varies considerably from 
cultural subgroup to cultural subgroup. The norm requirements of 
the statistical-frequency point of view become most discouraging 
when we realize that it is possible to conceive of a very large number 
of behavioral continua for any one subculture. 

There is also the point of view that what is normal and what is 
abnormal may be specified with respect to some sort of ideal; i.e., 
normal is considered to be what is desirable. Upon reflection it seems 
possible that the sources of such ideals of desirability may be quite 
numerous and would include political and ethical considerations and 
could change after every election. Of course, there are medical ideals. 
These seem to give no sure footing, however, because even in the 
aspects of medicine which seemingly are solidly founded and well- 
known change occurs, so that the ideal behavior for eating, bathing, 
sleeping, exercising, etc., seems to change from time to time. De- 
partures from ideal behavior even in the form of symptoms vary in 
the seriousness of their implications from area to area, depending 
upon the kinds of diseases that may be found to be endemic. 

There are those who would claim that, although one sees in the 
abnormal individual the same sort of behavior as is seen in the normal 
individual, somehow or other in the abnormal individual this behavior 
has a different meaning. The laws by which it is combined or regu- 

154 Kentucky Symposium 

lated are claimed to be different from the laws applicable to normal 
behavior. Presumably one can understand the essential abnormal 
nature of behavior only when one can see how the behavior in ques- 
tion is abnormally interrelated. This point of view, although super- 
ficially attractive, is rather disappointing when examined critically 
because it implies that one knows how behavior or behavioral mani- 
festations are normally interrelated. 

Concerning the absolute value of clinical judgments of the content 
of behavior, almost anyone who has attended numerous clinics may 
come away with the feeling that almost all human attributes and all 
human life circumstances have at one time or another been regarded 
as essential factors in the development of, or crucial evidence for the 
development of, a marked and tragic mental pathology. Neverthe- 
less, study of individuals who are not and have never been patients 
indicates marked prevalence of the handicaps, disappointments, pres- 
sures, conflicts, uncertainties, abuses, and disillusionments which are 
claimed in the clinic as sufficient explanation for the development of 
pathological conditions. 

Perhaps it is time to abandon the distinction between normal and 
abnormal; perhaps it is not truly relevant to the work of the clinical 
psychologist. To do this need not be construed as an abandonment 
of a goal of clinical psychologists. Obviously, it is not suggested that 
we should disregard the differences in behavior which we have at- 
tempted to relate to a concept of the abnormal. It is suggested in- 
stead that we give primary consideration to the differences in behavior 
from the particular standpoint of the conditions of their origin and 
change. We know that there are forms of behavior which are acutely 
distressing to the individual, and we know that also there are forms 
of behavior which may not be distressing to the individual but are 
acutely distressing to his associates. Perhaps the proper emphasis 
would be with the origin of these forms of behavior and with some 
concern as to why they cause distress. However, since there are so 
many ways in which an individual may be distressed or may distress 
others, it may be that what changes may not be practical. Perhaps 
our emphasis should be on how changes occur. 


The practical significance of description of current behavior states, 
especially abnormal behavior, is said to depend on the clinician's 
knowledge of personality. Such a point of view could be consistent 

Wittenborn 155 

with a behavior-change view of personality, but it obviously begs the 
question if personality is considered from the static-trait point of view. 

The research study of personality may be arbitrarily classified into 
two main approaches. There are students of personality who ap- 
proach their subject matter by investigating the manner in which 
various forms of behavior may be modified. The studies of interrela- 
tions between the appearance of forms of behavior and characteristics 
of the environment are often designed to give general understanding 
of the manner in which various forms of behavior may develop in the 
individual and the manner in which they may be changed. Unfortu- 
nately, when inferring modifiability of behavior many of these studies 
interrelate concurrent information descriptive of the individual with 
concurrent (or concurrently reported) information descriptive of the 

In a few instances the investigation of behavior modification has 
made good use of the two traditional forms for scientific investigation: 
i.e., the field or observational approach, where experiments of nature 
provide the data for drawing inferences or testing hypotheses; and the 
laboratory or analogical approach. 

The other main approach in personality research has to do with 
the determination of traits. These traits are inferences or inventions 
of persons interested in certain aspects of human behavior. The in- 
ventions of these traits may be based on a study of the interrelation- 
ships among different kinds of observations, but the observations are 
usually concurrent and with few exceptions the traits are static. It is 
likely that the person inventing traits will succeed in sampling in- 
dividuals and behavior which are relevant to his concern. He may 
invent traits on the basis of apparent relevance to political issues, per- 
sonal values, or the way in which a person behaves in a social situa- 
tion. All such inventions are arbitrary and serve only to provide aid 
for describing individuals in respects which may be of interest only 
to the persons who share the concern of the individual who invents 
the traits. It can scarcely be claimed that they have any basic sig- 
nificance that involves intrinsic, or ubiquitous, characteristics of 
human behavior. 


Although the concept of personality is vague and is approached by 
many psychologists with doubt and inconsistency, it is customary 
among most psychologists to approach the concept of ability with self- 
assurance. This self-assurance may, however, be ill founded, and it 

156 Kentucky Symposium 

s quite possible that many psychologists get an altogether unwar- 
anted feeling of confidence from the apparent objectivity and 
precision of the traditional statistical manipulations to which they 
;ubject the kinds of performance records gathered by test procedures. 

At present much of test development appears to be essentially tech- 
lological, more concerned with the technique of sampling -behavior 
han with behavior per se. It would appear that the broad relevance 
md general conditions for the behavior sampled by the test concern 
he test maker very little. He tends to express an almost exclusive 
nterest in the reliability of his test score and the correlational validity 
)f the test score with reference to some static criterion. This interest, 
dthough technically necessary, is more statistical than psychological 
n nature. If tests were used only in the context in which they were 
levised and predictions were made only for groups and with respect 
o the criteria with which they are validated, it is quite probable that 
he use of tests could proceed without serious abuses. Unfortunately, 
lowever, regardless of the manner in which the tests are made and 
:heir original purpose, they are soon applied to situations different 
: rom the original situation and to the evaluation of individuals. This 
latural effort to use the test in a general and flexible manner is ac- 
companied by a wide variety of abuses. 

What determines an individual's relative goodness with respect to a 
rest score may occasionally be the subject of some speculation, but 
kvith a few conspicuous exceptions it is rarely a subject of systematic 
exploration. The conditions of the individual's present and past 
jnder which a specified validity may be ascribed to the test are un- 
3xplored. It is obvious that many factors may be determining the 
relative test standing of an individual. 

Even when current determiners of the test score have been recog- 
aized by either the maker of the test or the person who uses it, the 
manner in which a person's test score is likely to change as a result 
af subsequent experiences is unknown. It seems probable that indi- 
viduals vary greatly in their susceptibility to change, and this may be 
an important limiting factor in test validity as we now determine it. 
[n college guidance work it is common to observe that a person's test 
score or a person's performance which the test presumes to anticipate 
may change greatly as a result of additional specific experience. Such 
changes can result from some insight that the individual acquires con- 
cerning his role in the situation, from new motivation, or from a more 
healthful affective reaction. So in many testing situations it is quite 

Wittenborn 157 

possible that an estimate as to whether or not the score is an optimal 
score or minimal score may be as important as the score itself. 

Test makers seemingly are little concerned in this, and those who 
use the tests apparently are satisfied to be guided by informal guesses. 
Until we recognize that in general we do not know to what situations 
the test may be safely generalized and to which ones it may not be 
safely generalized our practical testing procedures can lead us into 
serious unprotected errors. Perhaps we are justified in saying only 
that our tests are based on a sampling procedure, and the degree to 
which we can predict and otherwise generalize is determined entirely 
by the resemblance between the situation under which we test 
(sample) behavior and the situations to which we wish to generalize. 


The problem of change and the individual's susceptibility to change 
which has been suggested by our discussion of personality and ability 
may have a central relevance to clinical work, particularly since clini- 
cal work is concerned with therapy. Clinical psychologists in some 
situations conduct therapy themselves, in other situations they are 
expected to recommend appropriate kinds of therapy, and in still 
other situations they are expected to predict the outcome of possible 
therapies. All these responsibilities involve judging the individual's 
susceptibility to change as a result of certain experiences. This is 
ironic because one of the things about which we know the least is 
the conditions under which individuals change. 

There are changes which can be brought about in the individual 
with varying degrees of success but without any clear rationale to 
account for success or lack of success. For example, in a therapeutic 
situation it is often most important to give the individual some knowl- 
edge of his environment: what he can anticipate tomorrow or the next 
day, how he can expect people to react to him, what he can expect 
to find in various situations, etc. Just how these changes are brought 
about in the individual is not known, but procedures for conveying 
such information seem to be our heritage and include pantomime, 
the spoken word, and the printed page. 

Apparently knowledge concerning the individual himself can also 
be imparted to him, and this is often accompanied by desirable be- 
havioral changes. Certain special preconditions must usually be 
established before self-knowledge is imparted or gained. Unfortu- 
nately, it seems that the therapist must decide upon the nature of 

158 Kentucky Symposium 

these necessary preconditions by intuition and must evaluate them 
informally. After such evaluations of preconditions are made the 
question of whether the therapist conducts this teaching (or guides 
the learning) in a way which is more efficacious than any randomly 
selected way remains unanswered. As he conducts his therapy he 
has neither the benefit of an explicit rationale nor the support of em- 
pirical studies, Let us hasten to note that at the conclusion of any 
phase of therapy he may review his records and make an ex post 
facto application of some theoretical formulation to what has tran- 
spired, but can he use any current theory to plan and successfully 
anticipate the next phase of his therapeutic efforts? 


Many of the things that we do or say in relation to the words 
"abnormal/' "personality," "ability," and "therapy" lead us to both a 
practical and a conceptual emphasis on the role of behavior change. 
The concept of abnormal seems to have served us poorly, and we are 
tempted to turn to a point of view which does not emphasize a dis- 
tinction between normal and abnormal, but which seeks to understand 
the manner in which behavior and its personal and social values 
emerge and are modified. Our consideration of personality leads us 
to skepticism concerning the value of descriptions which are designed 
to serve no explicit purpose; our hope for progress in the study of 
personality includes an understanding of the conditions necessary 
for constancy and for change in behavior. Our remarks concerning 
ability and its assessment center about success and failure in predic- 
tion; such a discussion of prediction must turn to a further considera- 
tion of change in human behavior and of the problem of individual 
differences in susceptibility to change. Finally, when we discuss 
therapy, we find that our description of therapeutic change must in- 
volve the conditions for necessary prior changes, and we suspect 
that any plan of therapy must include evaluations ( either explicit or 
implicit) of the individual's susceptibility to diverse specified thera- 
peutic conditions. Thus it appears as we emerge from our brief 
discussion of concepts and practices of clinicians that our emphasis 
on change is greater than it was when we began. 

We become increasingly aware that the content of the behavior 
change which has our interest depends upon our values and our pur- 
poses, and they in turn we see to have but a regional and temporal 
significance. Accordingly, it becomes apparent that our conceptual, 

Wittenborn 159 

theoretical, and, if you please, scientific interest turns to the how of 
behavior change, and the content or the what of the change in be- 
havior is but an accident of time and place. An emphasis on the how 
of behavior change is not only an implicit expression of a hope but 
also an implicit application of a premise which is familiar in much of 
the work of psychology; specifically, it implies that change in be- 
havior, regardless of content, may be understood (i.e., may be ex- 
plained, predicted, and produced) by an employment of a single set 
of concepts. This premise implies that one psychology of behavior 
change may be sufficient. The antithesis to this premise would be 
either that no theory of behavior change is possible or that a differ- 
ent theory of behavior change is necessary for different behavioral 
contents. It would seem that a denial of the premise that one theory 
is sufficient for behavior change is an admission of a possible chaos 
in the study of human behavior, and it implies that psychology can 
be no more than a technology for sampling responses and for gen- 
eralizing from such a sample of responses to similar samples of re- 

If it is reasonable to think of the psychology of the clinician as the 
psychology of behavior change, we must acknowledge that we have 
been, and are, extraordinarily coy about studying behavior change. 
It is obvious that we have tarried long with descriptions of behavior 
states and have been obsessively concerned with the what or the con- 
tent of samples of concurrent behavior. If description of particular 
contents has been our purpose, we not only have often failed to say 
why we wish to describe (we have neglected to say how we hoped 
our description might be useful), but it can scarcely be claimed that 
we have done well in our descriptive efforts. Even our techniques 
of description seem to be content bound, and if there are broadly 
general considerations which can be applied to any problem of be- 
havioral description most of us would have difficulty in verbalizing 
these principles in any succinct and mutually acceptable manner. 

It is obvious that if we are to study behavior change we must de- 
scribe what is changing in order to have referents in our discussion 
of change per se. It seems, however, that the time has come when 
we must acknowledge the technological nature of our descriptive tasks 
and explicitly develop the principles of description per se and not 
be content to publish anecdotes (numerical or otherwise) of our ad- 
ventures in describing behavior. One wonders if the time has not 
come also to suspect that description without concern for change is 
likely to be an evasion of our practical and scientific responsibility. 

160 Kentucky Symposium 

Occasionally, one hears clinicians and other students of human be- 
havior say, "We must turn to learning theory." Although such state- 
ments imply an acknowledgment of the limited value of much of our 
descriptive research, they also imply that we have but to "turn" to 
learning theory to get our answers. As yet there is no reason for 
believing that learning theory has the answers, and it is possible that 
systematic and thoughtful observation of the changes in human be- 
havior under diverse conditions can contribute more to learning the- 
ory than learning theory can at present contribute to the work of the 
clinical psychologist. Clinical psychologists should not underesti- 
mate their opportunities for the careful description of behavior 
change and correlation of such behavior change with prior character- 
istics of an individual and with current features of his environment. 
Such studies can be conducted as procedures for examining hypo- 
thetical deductions from some recognized theory, or they can be 
frankly exploratory and possibly the basis for inferences. It seems, 
however, that the study of changes (naturally occurring, educational, 
therapeutic, or laboratory) is not only the urgent responsibility but 
also the golden opportunity of those who do research in human be- 
havior, particularly in clinical psychology. 


Ability, concept of, 155 

Abnormal, concept of, 153f. 

Adams, D. K., author, 66-79; 98, 105 

Adler, A., 115 

Ammons, R. B., author, 138-147 

Antonitis, J. J., 40, 53 

Anxiety, 99, llOf. 

effect of, on drive level, 18f. 
Anxiety reduction, 119-121, 124f. 
Appetite, 87 
Arnold, L. P., 59, 65 
Association test in proposed experiment 

on therapy, 126f. 
Aston, F. W., 73 
Attitude-interest, objective measurement 

of, 98 
Attitudes, 98, 100 

Babb, H., 34 
Baker, K. E., 35 
Baldwin, J. M., 130 
Berg, E., 30, 35 
Bersh, P. J., 40, 53 
Bertalanffy, L., 70 

Bimodality of scores on insoluble prob- 
lem, 59 
Block, J., 146f. 
Bohr, N., 73 

Bronfenbrenner, U., 76, 79 
Brown, C. J., 52 
Brown, U., 38, 53 
Burt, C., 109, 113 
Bush, R. R., 6, 20 
Butler, J. M., author, 114-128 
Butler, R. A., 43, 52 

Cattell, J. McK., 104 
Cattell, R. B., author, 91-113 

Clinical psychology, research trends in, 

Cohn, S. H., 81f., 90 

Cohn, S. M., 81f., 90 

Combination of factors in learning, ad- 
ditive or multiplicative, 14-16, 20 

Combs, A. W., 137 

Conditioning, laws of classical, 4, 14 
Hull's theory, 9 

Spence's emendation, 912 
instrumental, 15 

Conscience and feedback, 89 

Consciousness, an integrative mechan- 
ism, 84 

Constructs, hypothetical, 68, 72-74 

Cross, K. P., 113 

Curiosity, a drive, 39, 99 

Cybernetics, 84-87 

Diagnosis, 148 

Differences, individual, utilization of, in 

psychology, 104 
Dodson, J. D., 53 

Dollard, J., 7, 20f., 55f., 65, 119-124 
Drive level, 18f. 
Drive strength, 107 
Drives, exploration and curiosity, 39 

externally elicited, 39f. 

homeostatic, 37 

Eckstrand, G. A., 24, 35 
Edwards, W., 62f. 
Eglash, A., 65 
Ego theory, 87, 121-125 
Einstein, A., 69 
Ellen, P., 57, 59, 65 
English, H. B., 130-132, 137 
Ergs, 99, 107 



Errors, experiment on, 145f. 

postulates about, 143-145 
Estes, W. K., 6, 20 
Experiment, univariate, 108f. 
Exploration, a drive, 39 

in monkeys, 4752 
Eysenck, H. J, 94, 96, 113 

Factorization of increments, 107 

condition-response, 107 
Factors of personality, see Personality 
Fantasy, 88 
Farber, I. E., 20f. 
Feedback principle, 84f. 

conscience and, 89 

homeostasis and, 86 

negative and positive, 86f. 

rewards and punishments, 82, 85 
Feigl, H., 74 
Finger, F. W., 40, 53 
Fisher, R. A., 109 
Fitch, F. B., 147 
Fixated behavior in rats, 56-60 

possible causes of, 58 
Flynn, J. P., 40, 53 
Fonda, C. P., 126-128 
Foster, H., 35 

Freud, S., 90, 98, 114f., 138, 147 
Frustration produced by insoluble prob- 
lem, 59 

effect of, on subsequent learning, 59 

production of unlearned behaviors by, 

Gagne, R. M., 35 
Gately, M. J., 40, 53 
Generalization, response, 30, 122 

stimulus, 25, 122 
Gibson, E. J., 139, 147 
Glaser, N. M., 65 
Grant, D. A., 30, 35 
Grice, G. R., 20 
Gulliksen, H., 6, 20 
Guthrie, E. R., 8, 20, 23, 35, 51, 139 

Habit, concept of, 82f . 
Haigh, G. V., 126, 128 
Hall, M., 147 


Harlow, H. F., author, 36-53; 82, 99, 

107, 113, 130, 136 
Hebb, D. O., 72, 74f., 79 
Heistad, G. T., 126, 128 
Hilgard, E. R., 129, 137 
Holton, G., 69, 78 
Homeostasis, feedback and; 86 
Homeostatic drives, 37 
House, B. J., 40, 53 
Householder, A. S., 20 
Hovland, C. I., 25, 35, 147 
Hughes, R. H., 35 
Hull, C. L., 6-9, 11-16, 20f., 23, 35, 

87, 96, 104-106, 108, 113, 139, 

143, 147 

Humphreys, L. G., 56, 65 
Hypothetical constructs, 68, 72-74 

Information theory, 85 
Isomorphism, psychophysical, 71 

Janet, P., 114 
Jerome, E. A., 40, 53 
Johnson, H. M., 67, 79 

Keller, F. S., 40, 53 

Klee, J. B., 65 

Kohler, W., 39, 67, 69 

Koffka, K., 35, 139, 147 

Krech, D., 67-72, 74, 77, 79f., 131 

Landahl, H. D., 20 
Lawrence, D. H., 24, 34, 35 
Laws, psychological, 76f. 

of learning, 3-5 
Lazarus, R. S., 126, 128 
Learning, a new concept in psychology, 

anxiety, effect of, 18f. 

factors affecting, 61 

field organization in, 132 

laws of, 3-5, 76f. 

not a single process, 132 

perception a factor, 62, 135f. 

personality pattern, 134f. 

phenomena studied in the laboratory, 

practical nature of the concept, 131 

"real life" vs. artificial situations, 2 



Learning, see also Conditioning, The- 
ories of Learning 
Levine, L., 146f. 
Lewin, K., 76 
Lobotomy, mental changes in, 102 

MacCorquodale, K., 72, 74, 80 

Maier, N. R. F., author, 54-65; 104, 

108, 113 

Maladjusted persons, types of, 126 
Marquart, D. L, 59, 65 
Maxwell, J. Clerk, 69 
McClearn, G. E., 42, 53 
McDougall, W., 98 
McNemar, Q., 146f. 
Meehl, P. E., 72, 74, 80 
Mendel, G., 73 
Miles, R. C., 53 
Miller, A., 113 
Miller, N. E., 7, 20f., 55f., 65, 87, 90, 

119-124, 143, 147 
Miniature systems, 70 
Monkeys, discrimination problems, 42- 

manipulation in, 4042 

visual exploration, 4750 
Montague, E. K., 21 
Mosteller, F., 6, 20 
Mote, F. A., 40, 53 

Motivation, anxiety in motivation the- 
ory, 38, 52 

application of factor analysis, 98, 100 

curiosity and manipulation, 3942 

homeostatic, 37 
Mowrer, O. H., aulhor, 81-90; 38, 53, 

90, 94, 108, 113, 121, 143, 147 
Murray, H. A., 98 

Neurophysiology, use of, in psychologi- 
cal theory, 6, 69, 74f . 
Neurosis, feedback and, 89 
Neuroticism, CattelFs factor, 95f. 
Newton, I., 75 

Oscillatory inhibitory potential, 10-13 
Osgood, C. E., 85, 90 

P-technique, 100, 107 
Pavlov, I. P., 67, 80, 87 

Pearson, K., 115 
Pennock, L., 53 

Perception, a factor in learning, 62, 

Guthrie's treatment of, 23 

Hull's treatment of, 22 

stimulus-response theory applied to, 


Perin, C. T., 5, 21 
Perkins, D. T., 147 
Personality, concepts of, 154f. 

factors of (Cattell's), 95f., 102, 110 

research, five steps in, 93 
Petrie, A., 102, 113 

Physical, meaning of term, 66, 69, 71 
Pitts, W., 6, 21 
Planck, M., 69 
Position responses, see Fixated Behavior 

in rats 
Post, S., ]46 

Postulate system about errors, 143-145 
experiment to test deductions, 145f. 
Psychotherapy, as changing of behavior, 

example of research on, 126 

history of, 114-116 

learning theory and, 117120 

publication trends in, 152 

R-technique, 94, lOOf. 

Rashevsky, N., 6, 21 

Reasoning, Dollard and Miller's theory 

criticized, 55f. 

Reinforcement theory of learning, 8 
Research trends in clinical psychology, 


Response generalization, 30, 122 
Response hierarchy, 14 
Reward, as feedback, 82, 85 

in human learning, 111 
Reward-punishment ratios, effect of, on 

learning, 60f. 

Role-playing and story completion, 62 
Ross, R. T., 147 
Rutherford, E., 73 

Schoenfeld, W. M., 40, 53 
Self, see Ego theory 
Self-concept, 98 



Sentiment, 77-79, 98 

Shaffer, G. W., 126, 128 

Shannon, C. E., 85, 90 

Sherrington, C., 75 

Simon, C. W., 53 

Skinner, B. F., 108 

Snygg, D., author, 129-137 

Solem, A. R., 65 

Source traits, 97 

Spearman, C., 96, 109, 113 

Spence, K. W., author, 1-21; 35, 51, 96, 

103, 105f., 108, 113, 117, 129 
Sperry, T. W., 123, 128 
Stimulus generalization, 25, 122 
Story completion, 62 
Superego, 95 
Systems, miniature, 70 

Taylor, J. A., 21 
Tests, 156 

Theories of learning, aided by personal- 
ity study, 101 

assumptions underlying, 131 
contribution of to clinical psychology, 


Guthrie's theory, 8 
Hull's theory, 6, 8 

Spence's emendation, 8-20 
neglect of organism, 103 
Rashevsky's, theory, 6 
Tolman's theory, 5, 8 / 

Theory, defined as postulate system, 139 

its value, 138 
Thinking, 87f., 123 
Thompson, M., 21 
Thorndike, E. L., 82, 87, 108, 130 
Thurstone, L. L., 6, 21, 94, 109 
Tolman, E. C., 5, 8, 21, 3l, 98, 117, 

Traits, 155 

Uexkiill, J. von, 76 
Underwood, B. J., 35 

Variables, intervening, 5f., 72 
Hull's, 9-17, 104 
organismic, 104 

Wapner, S., 65 

Warren, H. C., 130 

Watson, J. B., 71, 87 

Weaver, W., 85, 90 

Whitehead, A. N., 68f., 80 

Wickens, D. D., author, 22-35; 53 

Wiener, N., 84, 86, 90 

Wilcoxon, H. C., 65 

Wilde, O., 93 

Williams, S. B., 5 

Wittenborn, J. R., author, 148-160; 96 

Wright, F. L., 75 

Wundt, W,, 104